Report No. 25793-MAG Madagascar Decentralization November 5, 2003 PREM I Southern Africa Africa Region Document of the World Bank Acknowledgements This report presents the results o f a close dialogue on decentralization between the World Bank and the Government o f Madagascar, in particular the Ministry o f Decentralization, as well as Madagascar's development partners (AD, Cooperation Franqaise, UNDP, EU, GTZ, Swiss Cooperation). The report builds on extensive research commissioned by the World Bank since 2001. This report has been preparedby a team consisting o f Wolfgang Fengler (Task Team Leader, AFTPl), Emanuela Galasso (DECRG), Aurelien Kmse (AFTPl), Frank-Borge Wietzke and (AFTPl). The team has received strong support and guidance from Jesko Hentschel (Lead Economist until September 2003, AFTPl). Dieudonnk Randriamanampisoa (Senior Economist, AFTP1) was instrumental in organizing the studies on local government finance, which were carried out by the consultancy Miara Mita and led by Mamy Ralaivelo. These commune finance studies were funded from the Norwegian Trust Fund (managed by Jacomina de Regt, AFTES). The peer reviewers were: Harold Alderman (AFTHD), James Fitz Ford (TUDUR), Jennie Litvack (MNSED), andDana Weist (PRMPS). The report was prepared under the supervision o f Philippe L e Houerou, Sector Manager for AFTP1. H e offered conceptual guidance, provided critical analytical advice, and ensured quality control and management support. Hafez Ghanem, Country Director, supported the process and provided overall guidance. Peter Moll, Jesko Hentschel, and Willem van Eeghen commented on earlier versions o f the paper. Assistance with editing, communication, and budgetmanagement was provided by Cecile Wodon (AFTPl). O n the Government side, the team would like to thank the officials in the Malagasy government as well as provincial, regional, and local authorities who provided important insights for this study. The team worked very closely with the Ministryo f Decentralization, the Ministry o f Budget and Finance, the Ministry o f Interior, the Ministry o f Health, and the Ministry o f Education. The team appreciated in particular the support received from Gabhy Rajaonesy (Director o f Studies and Transfer of Competencies) and Herilanto Raveloharison (Secretary General) inthe Ministryo f Decentralization. The results o f the study were presented on September 29, 2003, at a National Conference on Local Development in Antananarivo, which was organized by the Ministry o f Decentralization. The outputs o f the conference, which will be documented separately, have been included inthis study. TABLE OF CONTENTS Executive Summary ............................................................................................................. i 1 Introduction . ..................................................................................................................... 1 2 Today's Decentralization Agenda . and Institutional Determinants . ....5 THELEGAL FRAMEWORK................................................................................................... Challenges 5 TheStatus of Provinces .................................................................................. 6 TheStatus of Sub-Provincial Governments........................................................ 7 DECONCENTRATION AND DECENTRALIZATION .................................................................. 7 AdministrativeParallelism............................................................................. 7 Deconcentrated Service Delivery............................................................. 8 CONTROL SYSTEM ........................................................................................................... 12 Control over decentralized government................................................... 12 Control over deconcentrated line agencies................................................ 14 REGIONAL PLANNINGBODIES .......................................................................................... 14 TOWARDSMORESERVICE-ORIENTEDADMINISTRATION................................................ A 16 3 Fiscal Framework . ......................................................................................................... THEFISCALGAP............................................................................................................... 19 19 EXPENDITURES 21 Generalprinciples of expenditure decentralization and assignment......................... ................................................................................................................ 21 Madagascar's sub-national expenditures .................................................................. 23 Geographical allocation and budget execution ......................................................... 26 REVENUES ....................................................................................................................... 28 Challenges in sub-national revenue mobilization ...................................................... 28 Madagascar's revenue challenge............................................................................... 32 TRANSFERS ...................................................................................................................... 35 Filling the sub-national gap ....................................................................................... 35 Madagascar's transfer system.................................................................................... 38 IMPROVINGMADAGASCAR'S FISCALFRAMEWORK .......................................................... 42 4 Communes . ..................................................................................................................... 47 INSTITUTIONAL ARRANGEMENTS AT THE COMMUNELEVEL ............................................. 48 Communeinstitutions andfunctions .......................................................................... 48 Sub-communegovernments ........................................................................................ 50 MEASURING COMMUNEFINANCE: SAMPLEAND METHODOLOGY ..................................... 52 COMMUNENEEDSAND OVERALL FLOWS OF FUNDS ......................................................... 53 Needs: The localfinancing gap.................................................................................. 54 Composition of availablefinancing............................................................................ 55 COMMUNE REVENUES ...................................................................................................... 56 Local revenue assignment and collection................................................................... 56 Therural-urban divide ............................................................................................... 57 Rural revenueperformance and composition ............................................................ 59 COMMUNE EXPENDITURES ............................................................................................... 62 Urban versus rural expenditure ................................................................................. 62 Rural communes spending ........................................................................................ . 63 IMPROVINGSERVICE DELIVERY AT THE COMMUNELEVEL ............................................... 66 Annexes Annex 1:Matrix of Policy Recommendations.. ......................................................... 71 Annex 2: Distribution o f Ministerial Functions................................................,........ 72 Annex 3: Revenue Assignment inthe 2000 and 2001 Budget.. ....................................... 74 Annex 4: Deconcentration of Expenditures.. ..................,.............,,..,.......,............,,. 75 Annex 5: The LocalFinancing Gap Methodology.. .................................................... 78 Annex 6: The Representativeness o f the 232 Commune Sample.. .................................... 82 Annex 7: Local Government Organigram................................................................ 85 Annex 8: Local Government Revenue Assignments. .................................................,. 86 Annex 9: Bibliography...................................................................................... 88 Annex 10: Glossary o f French and Malagasy Terminology.. .......................................... 92 TABLES Table 2.1 Administrative parallelism 8 Table 2.2 Personneland budget functions inthe social sectors 12 Table 2.3 The control fi-amework 13 Table 3.1 Possible expenditure assignments by level o f Government 22 Table 3.2 Composition o f central government expenditures 24 Table 3.3 Possible revenue assignments by level of Govemment 29 Table 3.4 Revenue assignments inMadagascar 34 Table 3.5 Combining objectives and design 38 Table 3.6 Transfers to administrative levels (inmillion MGF) in2001 39 Table 3.7 Transfers to commune budgets (2001) 40 Table 4.1 Local financing gaps (US$ per capita and % o f total needs) 54 Table 4.2 Sources o f Financing (US$per capita) 56 Table 4.3 Revenue items and collection mechanism 57 Table 4.4 Revenue composition o f urban and rural communes (estimated) 58 Table 4.5 Ownrevenuebyrecovery mechanism 62 Table 4.6 Breakdown of expenditures inurban communes and communes inthe 63 greater Antananarivo area Table 4.7 Breakdown o frecurrent expenditures 64 Table 4.8 Breakdown of investments 65 Table 4.9 Functional classification - overview 66 CHARTS Chart 2.1 Madagascar's Territorial Administration 11 Chart 3.1 Madagascar's revenues inthe international context - Government revenues 20 as a share o f GDP (excluding grants) 1998 Chart 3.2 Madagascar's revenues inthe internationalcontext - Government revenues 20 as a share % o f GDP (excluding grants) 2001 Chart 3.3 Shares o frecurrent expenditures executed below the center 24 Chart 3.4 Social sector spending by administrative level 25 Chart 3.5 Central Government Budgetary Revenue 1998-2000 33 Chart 3.6 Delay o f unconditional transfers to rural communes (sample o f 15 41 communes) Chart 4.1 Institutional relationships at the local level 52 Chart 4.2 Data Sources 53 Chart 4.3 Gap per sector (US$ per capita) 55 Chart 4.4 Breakdown o f total local revenues 60 Chart 4.5 Breakdown o f total local taxes 61 Chart 4.6 Breakdown o f user fees 61 Chart 4.7 Breakdown o f administrative fees 61 Chart 4.8 Economic and functional expenditures o frural communes 63 BOXES Box 2.1 The Inter-communal association Fianarantsoa-Manakara 15 Box 3.1 Municipal Revenues in South Africa (2001) 32 Box 4.1 Local development planning (Plan Communal de Diveloppement, PCD) 49 Box 4.2 Managing local security issues inuncertain environments 51 Box 4.3 Case study o f the schools inthe school district o f Farafangana and 54 Andilamena Box 4.4 Social funds and decentralization -movingprogressively away from the 59 rural-urban divide? EXECUTIVESUMMARY 1. The democratic transition o f the early 1990s provided the context for decentralizing some powers to lower levels o f the Malagasy government. Communes became the focal point o f Madagascar's decentralization strategy and central ministries increased their local presence through administrative deconcentration. Since then, the decentralization process has been subject to intense political bargaining, leading to several strategic turns inthe reform process. 2. Today, there is substantial uncertainty about the future direction o f the decentralization process. This uncertainty has created tensions within and between various layers o f government. These tensions and uncertainties revolve around two main issues: first, the devolution o f authority and responsibility to lower levels o f government, although decided, was never fully implemented. At the provincial and commune levels, the central government continued to carry out many o f the executive functions o f decentralized institutions, even after these institutions hadbeen established. Second, the 2002 politicalcrisis and the subsequent change in government created a strategic vacuum in the country's decentralization policy. The new government de facto reversed the radical reform o f 1998, which established the autonomous provinces without officially announcing a new strategy. 3. N o w that Madagascar has emerged from the crisis, the government needs to take major decisions with respect to the future direction o f the decentralization process. Regardless o f which approach it will adopt, this will be a long and risky process. Madagascar's new government takes on the legacy of a far-reaching reform centered around the establishment o f six autonomous provinces. Given the tensions that arose during the 2002 crisis between the new government and these autonomous provinces, a major reorientation o f the decentralization process i s expected, and in many respects desirable. The election o f local governments - due to take place in November 2003 - will also provide an opportunity to clarify the roles and responsibilities o f communes, which are, in practice, the only effective level o f sub-national government. 4. Decentralization, inany form, will not be a magic bullet for overcoming Madagascar's development challenges, even though it has often been seen as a panacea for solving Madagascar's govemance problems. Inmany countries, especially in Africa, decentralization results from the failures o f a poorly performing centralized state, and in this context, decentralization does not present a recipe for automatic success. 5. As inmanyother poor developing countries, there are substantial structural constraints to decentralization in Madagascar. These include a large fiscal imbalance between the center and sub-national layers, lack o f capacity to carry out decentralized services, and a weak legal system. These constraints defacto exclude many o f the theoretical options for decentralization typical o f developed or transitional economies - in particular with respect to fiscal decentralization (e.g. sub-national borrowing, tax sharing, sophisticated performance-based transfer system). i Madagascar remains a very centralized country 6. Despite having embarked on several waves o f decentralization since the early 199Os, Madagascar remains a very centralized country, both administratively and fiscally. 7. Administrative centralization. Even after ten years o f decentralization policies, very few administrative functions have actually been decentralized. Policy planning, personnel management, and budgetingare still carried out by the central government. Even facility-level management has remained centralized - except in cases where donors work directly with the communities. Commune competencies are limited to some classic functions (administrative services, waste management) and some co-financing o f social services through conditional grants. Inpractice, deconcentrated agents still co-administer local government functions. 8. The deconcentrationo f services is slightly more advanced. The partial deconcentration o f social sector ministries has improved the outreach o f health and education services, but the discretion o f deconcentrated layers over personnel and budget management remains very limited. Most decisions are still taken at the central level, inthe capital Antananarivo. 9. Fiscal centralization. The centralization o f Madagascar's revenues is higher than in most other low-income countries for which data is available. The central government collects more than 97 percent o f total revenues. The only truly decentralized level, the communes, accounts for 2-3 percent o f total revenues, out o f which almost half i s attributable to urban communes. Rural communes collect on average only US$1per capita per annum. 10. The high degree o f revenue centralization and the minimal volume o f sub-national revenues are the result o f the twin impact o f (i)a very low overall revenue base representing 9.6 percent o f GDP (2001), and (ii) a disproportionately high share o f revenues items which cannot or should not be decentralized. Trade taxes, corporate income taxes, and VAT accounted for 78 percent o f total revenue in2001. 11. The centralization o f revenues i s mirrored by a very centralized system o f public expenditures, although some steps were taken to deconcentrate the execution o f expenditures in the late 1990s. In total, the central government manages and executes 88 percent o f total government expenditures. Deconcentrated expenditures amount to 10 percent, executed by deconcentrated line agencies o f the ministries, mostly in the social sectors. Recurrent expenditures are more deconcentrated than investments, which are almost completely executed from the center. Decentralized expenditure amounts to 4-5 percent o f the total, out o f which central government transfers to sub-national governments cover 3 percent, while only 1-2 percent i s attributable to sub-national governments' own revenues. Challenges and risks 12. Against the background o f Madagascar's structural constraints, it i s crucial to adopt a prudent and realistic approach: one that seeks to mitigate the risks o f the existing decentralization process and does not attempt to go beyond what can reasonably be achieved inthe medium term. Currently, the vacuum inMadagascar's decentralization strategy comes from the uncertainty surrounding the status of the autonomous provinces. The autonomous 11 .. provinces were introduced (together with the regions) with the 1998 amendments to the constitution. They were made effective with the provincial elections in December 2000 and the provincial executive's nomination inJune 2001. 13. The 1998 constitutional changes were radical because, for the first time, the center had prepared for the transfer o f substantial competencies to sub-national bodies - the autonomous provinces. However, with hindsight, it seems that the establishment o f autonomous provinces only resulted in creating even stronger parallelism between deconcentrated and decentralized administrative functions because the respective competencies o f the different levels o f government were not clearly defined. While some parallelism is normal and necessary in any country with some degree o f decentralization, the lack o f clear roles and responsibilities has created confusion and tension between "competing" decentralized and deconcentrated levels o f government. 14. Arguably, full implementation of the 1998 decentralization strategy would create substantial risks. It i s difficult to imagine that such a radical transfer o f competencies would be carried out smoothly given Madagascar's high degree o f centralization and the weak state o f the legal system. There i s a particular risk o f (i) undermining the autonomy o f communes and (ii) weakening service delivery and macro-economic stability: Undermining the autonomy of communes. Under the 1998 strategy o f inter- governmental relations, the communes would have fallen under the legislation and financial oversight o f the provinces. Provinces would have been able to exercise almost full discretion over communes' finances, including transfers (accounting for 75 percent of rural communes' budgetary revenues) and local revenue assignments. With such a system, there would have been a major risk that competencies formerly decentralized at the commune level would have been ccre-centralized'yat the province level. Communes would have lost discretion over their finances, and hence their capacity to operate inany real autonomous way. Weakening services delivery and macro-economic stability. Almost none o f the newly established provinces are financially viable. Under-financing o f the autonomous provinces creates risks for the delivery o f government services and for macro- economic stability. To finance even minimal levels o f services, all the provinces would have to depend on substantial intergovernmental transfers. While intergovernmental transfers are not harmhlper se, the magnitude o f needed transfers, the imbalance o f provincial finances, and the absence o f a regulatory framework would mean highrisks o f structural sub-national fiscal deficits and a fragmentation o f social services. The role of communes 15. The most notable achievement o f Madagascar's decentralization policies was the establishment o f 1,392 communes in 1995. Currently, communes represent the only effective and viable layer o f decentralized government. Yet, communes have been the main victim o f political bargaining and uncertainty, which explains why many o f their assigned competencies have not yet been effectively transferred to local governments. For instance, deconcentrated agents are continuing to collect the most important sources o f local revenue. 111 ... 16. While many urban communes gained from decentralization - in particular through decentralized and hence improved revenue collection - most rural communes still lack finance, competencies, and capacity. Commune finance i s characterized by a great divide between urban and rural revenue mobilization performance. Urban communes have increased their own revenues several fold since the mid-1990s. In contrast, rural communes' own revenues remain marginal both in absolute and in per-capita terms; central government transfers still represent 75 percent o f their budgetary resources. 17. This paper argues that communes should be at the core o f the government's decentralization strategy, but there should be no illusions as to what communes can reasonably achieve. Commune finance, including transfers, represents between 3 and 4 percent o f total government finance and even less in rural communes, where a large majority of Madagascar's poor live. Local services are mainly carried out by the deconcentrated administration (for recurrent expenditures) and donors (for investments). Rural communes' own financing is even inferior to contributions o f community groups, which provide a growing share o f local service finance. As a result, communes mainly depend on central government transfers, which typically arrive late (by six months or more) and sometimes not at all. 18. Similar to many other poor developing countries, Madagascar's rural communes systematically under-exploit taxes like the property tax. As a consequence, most rural communes have to rely on alternative sources o f local revenue. For instance, larger communes receive important shares o f revenue through the levy o f user fees for local infrastructure (markets, housing, tourist sites, etc.). However, smaller rural communes that do not have comparable infkastructure endowments must rely on economically more counterproductive and volatile revenue sources such as sales taxes on local primary goods. 19. Itmustbe stressed, however, that communes are muchmore efficient inthe collection o f local revenues than deconcentrated intermediaries. Although deconcentrated and centralized agents are formally charged with the collection o f the majority o f local taxes and fees, they only managed to collect 30 percent of local revenues in the communes sampled in this study. Incontrast, the few revenues that are directly perceived by the commune (mainly administrative fees and some agricultural taxes) account for almost 70 percent o f local governments' own income. AN AGENDA FORIMPLEMENTATION 20. GivenMadagascar's current institutional and fiscal context, implementing a policy o f radical decentralization would be difficult, unnecessary, and probably dangerous. In order to foster institutional stability, the government should consider suppressing or postponing the installation o f an intermediary level o f government. This report suggests that, for the time being, the establishment o f autonomous provinces should not be pursued further. The focus should be on strengthening communes, with inter-communal associations providing a flexible option for regional planning. Eventually, an intermediary level o f government could play an important role to bridge the distance betweenthe center and the 1,392 communes. Currently, the risks associated with pursuingthe installation o f autonomous provinces seem to outweigh the potential benefits. iv 21. Furthermore, there i s major room and urgent need for improvement within the existing framework o f decentralized and deconcentrated service delivery. The government could focus its service delivery strategy on (i)strengthening rural communes, the core unit o f decentralized governance, and (ii)improvements within the existing mechanisms of deconcentrated service delivery.` Strengthening rural communes 22. Communes are the only operational - and hence the most crucial - sub-national layer o f government. However, the 1998 reform and recent decrees risk sidelining the communes. The communes should again become the unit o f reference at the local level and be governed by a more certain andstable framework. Central government interferenceshould be limited as much as possible. Communes need to receive a realtransfer o f competencies to raise revenues and most importantly, they need to see an increase o f their transfer allocation. 23. Strengthening sub-commune structures, like Fokontany and Fokonolona, is laudable and important because sub-commune institutions canplay an important role incontributing to communal functions. Inparticular, strengthening sub-commune structures could help to better extend administrative services to remote villages. Insufficient procedures to obtain basic registry services (such as birth, death, andmarriage certificates) cause major social problems for poor families. For instance, ex-post registration for birth certificates are extremely cumbersome to receive and expensive as they would entail a court order. 24. However, the detachment o f sub-commune structures from the commune's authority - as it is currently envisaged - carries the risk o f underminingeffective provision o f communal services. With too strong a focus on deconcentrated command lines, the recent reform threatens to jeopardize the autonomy communes have obtained since the mid-1990s. This paper argues that it would be crucial, instead, to establish clear reporting relationships between sub-commune institutions and communes and to limit the role o f central government intermediaries (Sous-Prefet andDAA) to control and supervision fhctions. 25. While pro-poor targeting o f government resources is desirable in a country with very highinequality, administrative efficiency considerations call for simplicity and predictability inthe fiscal relations betweenthe center andthe periphery. Therefore, inorder to improve the financial position o f communes, the initial focus should be on basic improvements and good "housekeeping." 26. First, the responsibility o f collecting the communes' own revenues should rest solely on the communes. Until now, three deconcentrated (DAA, Treasury, DGI) and two local (Commune and Fokonatany) institutions are involved in collecting local taxes. As expected from accountability principles and as demonstrated by field studies, communes are more efficient incollecting their own revenues than are deconcentrated intermediaries. 27. Second, communes suffer from the unpredictability with which transfers are executed. The foremost priority is to ensure predictable disbursements within a scheduled rhythm (e.g. every six months). Given that, on average, 75 percent o f rural communes' financing depends I See Annex 1for a compilation of all of this report's short- and medium-termrecommendations. V on transfers, predictable disbursements would allow communes to actually plan their expenditures and stop the existing practice o f ad-hoc budgeting. 28. Third, transfers to rural communes should be increased to help guarantee that rural communes have the minimum resources to function better. Rural communes should get at least the same per capita allocation as urban communes (FMG 8,800), which would also remove the artificial and regressive separation between them. The increase o f transfers to rural communes could be fiscally neutral ifthe transfers to the districts, which have so far not translated inthe equivalent local investments, were permanently suppressed. Improve service delivery within the deconcentratedframework before decentralizingfurther 29. Madagascar's recent decentralization process has been characterized by several radical turns. This has created uncertaintyandtensions at andbetweendifferent administrative levels. Given that the central administration and its deconcentrated outposts still hold the key to Madagascar's service delivery system, and taking into account the substantial risks associated with rapidly introducing additional layers o f sub-national government, this report suggests ways to make the existing set-up more efficient before decentralizing further. This would also foster the stability o f the communes. When a more stable institutional framework i s inplace, Madagascar will be in a position to gradually decentralize further and introduce an additional sub-national layer. 30. With the 1998 reforms, an already weak system o f control and audit was further challenged by the magnitude o f transfers to the provincial level and by the poorly defined command chains. The gradual transfer o f competencies to the deconcentrated level recommended in this study would require an upgrade o f control capacity and overall co- ordination. A possible tool for achieving such control is the integrated budget management system discussed under the new institutional reform project. An improved control, audit, and budget management system would also constitute an important pillar for decentralizing institutional responsibilities. 31. Inthe same context, the government needs to clarify competencies for the monitoring and evaluation o f key public services. The most important reforms require the creation of clear reporting rights between statistical services at the province and central levels, as well as an effort to harmonize data collection procedures. In addition, adequate investments are needed to analyze, treat, and verify data at the province and district levels, and to allow for better coordination mechanisms for information sharing with central statistics departments andbetween sectors. 32. To further strengthen the outreach o f central ministries and speed up budget execution, key services should be further deconcentrated, with a transfer o f technical operational power to the province level. Under this system, the deconcentrated treasury should also collect the provincial revenues foreseen inthe budget (particularly business and automobile taxes). 33. Given that districts represent the core unit for deconcentrated social service delivery and also arguably the optimal unit for poverty targeting, current social expenditure could be better targeted towards poorer areas. Better targeting to poorer districts could be achieved by giving a heavier weight to more remote andpoor districts. vi 1. INTRODUCTION 1.1 Madagascar i s recovering from a deep economic crisis, which stems from a political crisis in2002. Immediately after the crisis, the priority was to mitigate the impact on the poor and to jumpstart the economy. Madagascar's post-crisis development objectives are twofold: bringgrowth back - at least - to pre-crisis levels and improve the distribution mechanism to increase the poverty impact of growth and public services. This study i s an attempt to contribute to the second objective. It focuses on service delivery mechanisms and on options for financing these services. 1.2 Madagascar's pre-crisis growth helped raise living standards in the cities. However, most o f the country's poor live inthe countryside, and the fruits o f growth didnot reach them. A recent household data survey demonstrates that duringMadagascar's recent growth period (from 1997 to 2001), the share o f people living inpoverty remained broadly unchanged. The share o f the population below the poverty line was reduced from 71 percent to 69 percent, while the share o f the extreme poor didnot change at all. 1.3 Poor governance i s often considered to be the main reason why the rural poor have not benefited from growth and public services. The administration has been unable to effectively deliver core services in the areas critical for Madagascar's development (in particular infrastructure, social services, and agriculture). Similarly, rules and regulations are not followed or systematically applied, be it with respect to administrative or to legal decisions. 1.4 Decentralization has often been thought o f as a key instrument for improving governance inMadagascar. The challenge o f decentralization - and its complexity -is rooted in its many conflicting objectives and features. It also calls for tough tradeoffs that only a deep understandingo f local conditions can inform. On one hand, lower levels o f governments tendto have an information advantage: they are able to identify cheaper and more appropriate ways o f providing public services. On the other hand, decentralization is very difficult to carry out in poor countries like Madagascar, which lack the required financial and administrative foundations, as well as the capacity. Economies o f scale, important geographical disparities, and the desire to guarantee national unity all make the case for less rather than more decentralization. Defi. ition 1.5 Decentralizationcovers a broad range o f concepts, but it always deals with the transfer o f authority and responsibility from the central govemment to subordinate or quasi- independent government organizations or the private sector.2 Decentralization can have three dimensions: administrative decentralization, political decentralization, or fiscal See Rondinelli, "What is Decentralization?", in : Decentralization Briefing Notes edited by: Jennie Litvack and Jessica Seddon, World Bank Institute, 1999. 1 decentralization. Administrative decentralization, can take three major forms: (i)deconcentration, (ii) delegation, and (iii)devolution. (i) Deconcentration i s the transfer o f clearly specified decision-making, financing, and management functions to local line agencies, which depend directly on central government ministries. (ii) Delegation is the transfer of decision-making authority to semi-autonomous organizations, which are not completely controlled by the central government, but ultimately accountable to it. Examples include public enterprises, as well as housing andtransport authorities. (iii)Devolution is the transfer modality that is probably nearest to the general understanding o f decentralization. It includes the transfer o f authority and decision- making power to legally and politically autonomous sub-national govemments, inthe areas o f public finance and management. Devolution and decentralization are used interchangeably inthis report. Objective and challenges of this study 1.6 The objective o f this decentralization study is to provide analysis and policy advice to the new Malagasy government on how to proceed with the decentralization strategy it inherited from the previous government. The end o fthe post-crisis emergency recovery period creates the opportunity to correct the weaknesses o f the previous strategy while building on the exiting achievements. The paper specifically aims to (i) analyze the institutional and fiscal context o f decentralization and (ii) present the resulting challenges for service delivery and financing - with a particular focus on local govemments. Benefiting from primary data from several hundred local governments, this study aims at proving recommendations o f how to improve the functioning o f communes inconsidering the institutional and fiscal parameters. 1.7 Madagascar's changing governments and decentralization strategies presented the main challenge in carrying-out this work. The study operated ina context o f uncertainty about the nature and direction o f successive govemments' decentralization strategies. Even before the crisis o f 2002, the decentralization process hadbecome a moving target, inparticular with respect to the status and role o fthe autonomous provinces. Contentsand structure 1.8 Decentralization has so many dimensions - political, administrative, sectoral, and fiscal to name only a few - that decentralization studies necessarily need to limit themselves. At the same time, there is often a danger that sectoral specialists focus on their area o f specialization only. This "compartmentalization" o f decentralization also explains why the assessments o f decentralization processes in developing country wary so much: some grass- roots oriented practitioners consider it a blessing to move toward local empowerment, while many macroeconomists often fear the fiscally irresponsible behavior o f sub-national governments. This i s why a broad perspective i s needed to gain a balanced understanding o f decentralizationprocesses. 1.9 This study attempts to take such a broad approach by explaining the political, institutional, and fiscal determinants that have been influencing Madagascar's decentralization process, and which will continue to do so in the future. Based on an 2 institutional and fiscal analysis, the study presents an in-depth analysis o f communes - currently the only truly decentralized level o f government, in place for two electoral cycles. Thus, the study is structured along three pillars: (i) Politicalandinstitutionaldeterminants Chapter 2 provides the political and institutional background needed to better understand the decentralization context o f today. The chapter analyzes the historical forces o f centralization, which date back to pre-colonial times and translate into today's dominant role o f the central government administration - despite attempts to decentralize since the mid-1990s. The chapter then takes a closer look at the existing administrative set-up to explain functions and responsibilities at different levels o f government. To this end, it looks at the regulatory gaps and inconsistencies in the legal framework, the parallelism between decentralized and deconcentrated layers, and the emerging role o f inter- communal associations and informalregional planning bodies. (ii) Fiscalframework The salient feature o f fiscal decentralization in most poor countries i s the fiscal gap. Conditions for fiscal decentralization in Madagascar are less conducive than the in average poor or Afi-ican countries. Based on the concept o f the sub-national fiscal gap, chapter 3 analyzes structural constraints to fiscal decentralization in Madagascar, and in particular the dominance o f revenue items which cannot be decentralized. Given the low degree o f fiscal decentralization, the chapter takes a closer look at expenditure deconcentration and provides a more in-depth analysis o f deconcentration in the health and education sectors. The last part o f the chapter evaluates Madagascar's intergovernmental transfers, which represent the most important source o f local government financing. (iii) Communes Communes are the lowest and institutionally most advanced level o f sub-national government inMadagascar. They are also the main focus o f this study. The fourth chapter describes the institutional dynamics at the local level to provide the ground for a thorough analysis o f local government finance. The chapter provides an estimate o f local needs, and analyzes revenue and expenditure pattems based on budget data o f a large sample o f rural communes. 1.10 Given the natural overlaps between the three chapters some repetition and cross- referencing is unavoidable. For instance the section on commune revenues could be covered inthe fiscal as well as inthe commune chapter. As a general rule, broadtrends and aggregate results are presented inthe first two chapters. The last chapter then presents the details o f the institutional dynamics and fiscal pattems o f communes. 1.11 This study does not focus on areas which have been or will be addressed in related World Bank studies such as the education and health sector studies, the Country Financial and Accountability Assessment (CFAA), or the Public Expenditure Review (PER). Instead, this study concentrates on Madagascar's service delivery performance from the viewpoint o f local governments. Tracking surveys and a deeper analysis o f the institutional set-up at the health and education district administration level would complement the analysis inthis study. These background studies have been commissioned under the upcoming PER and the health sector study. 2. TODAY'S DECENTRALIZATIONAGENDA - CHALLENGESAND INSTITUTIONAL DETERMINANTS 2.1 Malagasy territorial administration comprises five main levels: provinces, regions, districts, communes, and villages. The establishment o f decentralized governments at the commune level (1995) and at the provincial level (2001) challenged the dominant role o f the central administration. The regions, although they were also granted constitutional status in 1998, have not been established yet. Districts, incontrast, are the key level o f deconcentrated social service delivery; they represent the most important interface between the central administration and the service facility. 2.2 Although these different levels have been defined inbroad terms, the exact extent o f their competencies and the nature o f their inter-relations remain uncertain. The 2002 political crisis and the subsequent change in government have created a strategic vacuum in the country's decentralization policy, which causes uncertainties and tensions at many levels. While the new government de facto reversed the prevailing policy based on autonomous provinces, it has not yet announced a new decentralization strategy. N o w that the country is beginning to recover from the crisis, the government needs to address this strategic vacuum and take decisions with respect to future institutional and administrative relationships in the country. These institutional dimensions include (i) legal framework, (ii) relationship the the between deconcentrated and decentralized administrations, (iii) the control system, and (iv) regional planning bodies. THELEGALFRAMEWORK 2.3 Withthe 1998 constitution and accompanying by-laws inplace, Madagascar embarked on a process o f asymmetric decentralization, which entailed the progressive transfer o f selected competencies to sub-national levels o f governments. Asymmetric decentralization reforms have beenpracticed inmany industrialized countries (United States, Spain), and they are also believed to be particularly suitable to developing countries, notably because they allow for a gradual adjustment o f sub-national capacities to newly assigned competencies. 2.4 However, this asymmetric approach i s also conducive to excessive political bargaining and to attempts by central authorities to block meaningful transfers o f competencies and resources. In Madagascar, these dangers were reinforced by an unclear legal framework that left uncertain the status ofprovinces and sub-provincial governments. 5 The status of provinces 2.5 The 1998 constitutional framework, with its strong focus on autonomous provinces, places the country at a crossroad from a unitary to a regionalist state.3 Even before the recent political crisis broke out into open opposition between central government and autonomous provinces, this new legislation created strong uncertainties about the future o f state-province relations. 2.6 The 1998 constitution formally grants strong legal and political autonomy to provincial governments, including the right to establish "provincial constitutions" (lois organiques) covering the areas o f finance and control. It also gives them the right to determine the functions as well as the fiscal, administrative, and control frameworks o f lower level governments. In addition provinces were assigned the right to negotiate their competencies directly with central government in a negotiating forum, the Confe`rence Inter- Provinciale (cIP).~ 2.7 At the same time, the constitution and by-laws contain a number o f regulatory gaps, which could severely limit the autonomy o f provinces. Unlike communes and regions, which enjoy the status o f "decentralized territorial entities" (Collectivite`s territoriales de`centralise`es), provinces are more broadly defined as regular "legal public entities" (Art 126, constitution). Inaddition to the imprecise definition o f their status, provinces have never been attributed financial and administrative competencies: 0 Provincial competencies will be negotiated in the CIP between the central and provincial governments resulting in substantial planning uncertainty. The constitution assigns strong discretion to the central government (represented by the presidency) to determine the timing and procedures under which the CIP is to take place. Under the previous government, the CIP's status already generated strong uncertainties about the expected duration and modalities o f the negotiation pro~ess.~ 0 Resources, such as budgets, equipment, or staff, would be put "at the disposal" (mis h la disposition) o f provincial governments without further clarification on whether provinces will have the right to determine budget allocations or to hire and fire civil servanh6 Madagascar has traditionally followed French-Napoleonic principles o f unitarian statehood. However, the degree o f legal autonomy and political discretion assigned to the six autonomous provinces by the 1998 constitution are more consistent with federalist or regionalist principles o f state organization (such as those o f Spain or Italy). Ifimplemented, these new regulations could open the door for legal-political arrangements which would be very difficult to absorb ina unitary state framework. According to the constitution, the areas that would remain under central control are: foreign relations, defense, national security, macro-economic and monetary policy, customs, human rights, as well as sectors "of strategic national interest" (Art 135 constitution). cf. Art 139, 150 o f the constitution. The constitution does not make provisions for a decentralized civil service code. However, provinces have the right to exercise the civil service code indelegation from the central government (loi organique 2000-16, Art. 17, 18). 6 The status of sub-provincialgovernments 2.8 Sub-provincial governments are equally affected by regulatory gaps in the current legal framework. If the 1998 constitution had been implemented, communes and regions would have lost most o f the administrative and financial autonomy, which they had been granted in 1994. Only the executive and legislative organs as well as the names and demarcation o f communes and regions are protected by constitutional law. In contrast, competencies, control mechanisms, revenue assignments, and transfer systems to lower level governments are placed directly under the legislation and oversight o f the province^.^ 2.9 Unless complemented by clear by-laws, these regulations create strong uncertainties for the future development o f sub-provincial governments. In the case o f communes, these uncertainties put at risk hard earned advances in the areas o f local revenue collection, financial management, and local service delivery. Given the inactivity o f the provincial legislative bodies after the recent crisis, communes today operate defacto under the laws o f 1994-95.8 DECONCENTRATION DECENTRALIZATION AND Administrativeparallelism 2.10 Since the decentralization reforms o f the mid-1990s, Madagascar's administrative set- up has been characterized by a strong parallelism between deconcentrated and decentralized functions across many levels (see Table 2.1). While some parallelism i s normal and necessary in any country with some degree of decentralization, it has been particularly extreme in Madagascar as a result of the incomplete and often inconsistent implementation o f decentralizationpolicies over recent years. 2.11 This parallelism between decentralized and deconcentrated command lines affects the implementation o f decentralization reforms as well as public service delivery in general. Administrative parallelism exists at the (i) central and (ii) sub-national levels: (i)At the central level, the Malagasy decentralization policy suffers from unclear assignment o f responsibilities among central government institutions with respect to decentralization policy. Currently, the responsibility for monitoring and managing the decentralizationprocess is divided between the Ministryo f Interior (control systems), the Ministryof Finance and Budget (revenue assignments), and a newly createdMinistryof Decentralization(monitoring provinces and communes, allocating transfers). (i) thesub-nationallevel,thelackofclearrolesandresponsibilitieshascreatedsome At confusion and tension between decentralized and deconcentrated levels o f go~ernment.~ At the provincial and the commune levels, deconcentrated agents have continued to carry Art 126, constitution and Loi organique 2000-16, Art 13. Names and demarcations o f communes can only be changed through approval by the cabinet (Conseil des Ministres). Regarding the institutional arrangements at the commune level, see Loi organique 2000-16, art 15,38,50 and chapter 4. Laws 1994-01, 1994-06 to 1994-08, 1995-005. On the challenges o f "co-administration" o f deconcentrated and decentralized service delivery structures, see Gillson / Mills 1995 and World Bank 1999. 7 out executive functions o f decentralized entities, even after sub-national governments were formally put inplace. Table 2.1: Administrativeparallelism Levels Decentralized Deconcentrated Status and competence institution command lines Center Ministryof MinistryofInterior The Ministry ofDecentralization is attached to the presidency and Decentralization responsible for the implementation and monitoring o f decentralization policy. Sector ministries The Ministry of Interior is in charge o f the deconcentrated administration and public security. The line ministries are responsible for the implementation o f sector policies and deconcentrated service delivery. 6 Provincial Pre'sidentde la Provincial councils elected inDecember 2000, govemors elected inJune Provinces Council, De'ligation Spe'ciale 2001. Responsibilities to be determined b y the inter-provincial conference. Govemor ww Since the political crisis o f 2002 and the subsequent change o f govemment, [suppressed since [previously: the decentralized institutions o f the province have been suppressed. The future July 20021 DPlPguk Ge'ne'raldu role o f autonomous provinces remains uncertain. Gouvernement.DGGl 28 Not yet set-up; Prefect The regions are identical to the prefectures. Regions Planned for under Main competence: developmental planning (to be taken over from the the constitution districts). 111Districts Suppressed Under-Prefect Electoral district for legislative election. (Sous-Pre'fet) Administrative unit for school districts (CISCO) andhealth districts (SSD). [former Under-Prefect carries out legal control. cantons] ComitPLocal de DPveloppement(CLD) The CLD is bringing together all deconcentrated services o f the district and headed by the deputyo f the district (Dipute? the mayors o f the communes. It decides on the allocation o f a central and by the Under-Prefect (Sous-Prifet) govemment transfer targeted at investments in the communes. 1,392 Commune council; De'le'gukAdministratif 1,347 rural and 45 urban communes elected their mayors for the first time Communes Mayor d 'Arrondissement inNovember 1995 and subsequently ina4-year cycle; some o fthe larger urban communes are identical to the districts. Deconcentrated administration i s represented by the DAA, who is appointed by the under-prefect. H e carries out legal control o f communes, but also fulfils executive functions inthe commune, like revenue collection. 4pprox. 14,000 President de Fokontany (PdF) Fokontany (FKT) were the administrative unit o f the Second Republic villages (1975-1992). They are further subdivided into approximately 40,000 (Fokontany) Fokonolona (FKL).The status o f FKTand FKL inrelation to decentralized and deconcentrated administrationremains unclear; Fokontany/Fokonolona (i) maintainpublic security and regulated formal anc customary law, (ii) health and education campaigns, and (iii) the support assist communes to monitor the village population and manage the Etat Civil. Deconcentrated service delivery 2.12 The central government administration and its deconcentrated outposts are still key to Madagascar's service delivery system. Deconcentrated line agencies,continue to exist at all levels o f government. The ministries o f Interior and Finance/Budget maintain an extended network o f deconcentrated agents at all administrative levels. The sector ministries, in 8 particular Health and Education, have a deconcentrated base in each district (see Table 2.1. and Chart 2.1). 2.13 The Ministw o f Interior officially represents the central government at all administrative levels through its network o f prefects, under-prefects, and "districts delegates" (De`lkguks Administratifs d 'Arrondissement - DAA). The head o f each jurisdiction is responsible for organizing elections, assuming local police and security functions, and ensuring legal control over commune governments. In many cases, representatives o f the Ministry o f Interior even continue to carry out administrative functions that should be devolved to communes and other sub-national governments, such as the maintenance o f cattle registries or the monitoring and coordination o f administrative units at the village level. 2.14 The Ministn,o f Finance (and Budget) continues to play a strong role inlocal revenue collection, including the administration and collection o f sub-national taxes, through its deconcentrated line agencies. Line agencies are represented at each o f the five sub-national levels by a hierarchical chain of Treasury agents and by tax collectors (see Chart 2.1). The structure o f the Finance Ministry also includes the Circonscription FinanciBre (CF). The domain o f each CF is slightly larger than the region and it i s characterized by the presence o f the Centre Fiscal, which i s the focal point for central revenues' collection. The CF also serves as a base for the National State Inspection (ContrGle des De`penses Engage`es - CDE - see Chapter 3). The ministries o f Budget and Finance were merged in 2002 and the process o f consolidating sub-national administrative units is still ongoing. 2.15 Among the sector ministries, the Ministries o f Education and Health (MINESEB and MINISAN)" stand out as they (i) represent the most deconcentrated ministries with a growing network o f deconcentrated agents, and (ii)account for about 20 percent o f government expenditures : 0 The province level administration consists o f a light coordination structure. The province head office (Direction Inter-re`gionale) is charged with the execution o f statistical monitoring functions, as well as the technical, administrative, and financial (education sector only) oversight over district administrations, but does not hold any significant responsibility for budget management or service delivery. 0 The district administrations' are the core unit o f social service delivery. They manage recurrent non-salary budgets for local primary schools and primary health facilities, as well as salary budgets for locally employed service staff (i.e. teachers "hors-solde "). They also carry out local procurements and provide technical oversight over service facilities. 0 At the local level, the education administration i s further supported by local agents o f the Zones Administratives Pe`dagogiques (ZAP),who cover almost all primary schools within a commune. Community-based initiatives, often supported by donor projects, loMINISEB : Ministbre de 1'Enseignement Secondaire et Education de Base; MINISAN: Ministbre de la SantC. See Annex 2 for a more detailed list of the distribution o f functions across administrative levels for the education, health, agriculture, and transport ministries. `ICirconscriptions Scolaires (CISCO) in the education sector and Service Sanitaire de District (SSD) in the health sector. 9 complement the deconcentrated administration. In locally financed schools, parent- teacher associations (FRAM and FAF) manage non-salary recurrent expenditures and acquired the right to hire and fire teachers. In the health sector, a cost recovery mechanism for drugs was introduced. Revolving funds are partly managed by district administrations and local NGOs (in the case o f larger district hospitals). Horizontal control is exercised by local user committees. 2.16 Despite efforts to deconcentrate tasks to local line agencies, most administrative functions of the health and education ministries remain located at the center. Madagascar's degree o f administrative centralization can be analyzed in three areas which are potential areas for sectoral decentralization: (i) budget programming, (ii) execution, and (iii) budget civil service management (see table 2.2.). li)Budgetprogramming: Sub-nationalcapacitiesfortechnicalmonitoringandevaluation are relatively well developed because M&E hnctions were delegated to provincial line agencies in both sectors. However, neither provincial nor district administrations have significant experience with financial planning as most budget allocations are traditionally decided centrally between the Ministry o f Finance and the responsible line ministries. In the whole system there is little experience with output or performance oriented budget programming as most budgets are based on past allocations and as budget increases are typically distributed evenly between districts. (ii) execution: Lineagencieshaverelativelylittleexperiencewithdeconcentrated Budget budget execution. Where this capacity does exist, it i s mostly located at the district, rather than the province level. Investment and most administrative and salary recurrent expenditures are concentrated at the center. Only non-salary recurrent expenditures for primary service facilities are deconcentrated at the district level.12 The heads of district administrations (CISCO and the SSD) distribute the funds across the schools and health facilities within each district. Two thirds o f these funds are assigned to the purchase o f services (mainly current maintenance), while the rest (purchase o f goods) i s equally dividedbetween administrative and school supplies. Inprinciple, there is some autonomy inthe way districts allocate the funds. [iii)Management ofthe civil service: Madagascar's civil service code remains very centralized. Hiring, firing, and sanctioning rights are controlled centrally by the responsible line ministry. Province administrations (Directions Inter-rkgionales) and districts have very limited discretion over personnel and budget management, and they have a limited role to (re)allocate staff within their respective geographical levels. However, particularly in the education sector, district administrations have some experience with the management o f contractual staff that are employed directly by district administrations, beneficiary groups, and private service providers (see Table 2.2). However, in the health sector, the number o f contractual staff has been considerably reduced following a large regularization campaign financed through HIPC funds in2001. l2 Investment budgets and salaries are almost fully executed at the central level. See chapter 3 o n expenditures for the execution rates. 10 E; .I Y E c, v) .L -1 d 2 m Q .L L Y 0 .L L L c, L u 0 Q Table 2.2: Personnel and budget functions inthe social sectors Budget planning Budpet execution " Personnel management Center Lineministries Ministries for recurrent Hiring, firing, and determine global administrative expenditures sanctioning o f staff budgetenvelopes and the whole investment budget Deconcentrated apencies: Payment o f salaries for Allocation o f staff 0 Province level: Directions teachers and doctors within province inter-rkgionales 0 District level: CISCOs, Allocation o f Execution o f recurrent non- Allocations o f staff SSDs budget across salary expenditures within districts. Hiring facilities (maintenance, materials) and firing o f locally employed teachers (education sector) Decentralizedagencies: Execution o f conditional 0 Communes grants on healtWeducation Co-financing FRAM- managed schools CONTROL SYSTEM 2.17 Inadecentralized environment, controlfhctions canbeanalyzedalongthree dimensions: (i) The object of control: Control can be exercised over decentralized governments or deconcentrated line agencies. (ii)horizontal Control mechanisms: vertical control through central inspection agencies or control by the legislature and/or beneficiaries (mostly in the case o f decentralized governments). (iii)Typeof control:Legalcontroloverproceduresanddecreesorfinancialcontrol over the use o f public funds. 2.18 The following paragraphs focuses on the consequences o f the current regulatory vacuum on the control framework for decentralized governments. It also includes a brief description o f the existing control system for deconcentrated line agencies. A more thorough overview o f the Malagasy control system will be provided by a forthcoming CFAA. Control over decentralized government 2.19 For decentralized governments, the control framework's effectiveness i s particularly affected by the current legal and regulatory uncertainties. With the constitutional changes o f 1998, the control framework for sub-national governments was (re-)defined in the context o f the transfer of competencies to provincial governments. Communes and regions were considered sub-units o f the provinces, and accordingly the provinces were charged with defining control functions for sub-provincial governments. 12 Now, the defacto destitution o f the provinces has halted these legal reform initiatives andcreated a vacuum inthe reform o fthe control framework. 2.20 Vertical control. Like in many other countries with ongoing decentralization reforms, vertical control mechanisms have been weakened by the creation o f communes andprovincial governments. Since the mid-199Os, when decentralization began, the only systematic control o f local governments is inthe area o f legal control (carried-out by the Sous-prkfet with the assistance o f the DAA). Arrangements for financial control differ according to commune types (see table 2.3.). External financial control only exists for larger communes (exercised through a network o f deconcentrated Treasury officials) but not for the majority o f smaller rural communes. Financial audits can be carried-out through the internal audit body (IGE). 2.21 At the current stage, most vertical control mechanisms are ineffective for a number reasons. Legal control is hampered by the inconsistencies and uncertainties in the legal framework governing commune rights and responsibilities. Financial control functions and audits are rarely executed due to the lack o f capacity at the local level and the generally small levels o f expenditure that are managed by local governments. In addition, the absence o f deconcentrated administrative and financial courts complicates sanctioning and conflict resolution between local governments and control organs, and renders the control system vulnerable to political interferences. I Table 2.3: The control framework Legal control IFinancial control IFinancial control (accounting) (spending decisions) Urbancommunes PrCfet/ Sous-prefet Public Treasury Commune Council Rural communes Sous PrCfet / DAA Public Treasury (local Commune Council (premih-ecatdgorie) delegate) Rural communes Sous PrCfet / DAA Intemal control Commune Council (deuxidme catdgorie) executed the by commune itself 2.22 Horizontal control. Giventhe weaknesses inthe current vertical control system, there is an urgent need to strengthen upward control and horizontal accountability mechanisms. Formal horizontal control on spending decisions and accounting practices (Contr6Zed 'opportunitb et rkgdaritk) should theoretically be exercised by the commune council. However, in practice, this control i s often limited by insufficient training o f council members andor political conflicts between council members and mayors. The development o f more informal horizontal accountability mechanisms, such as through civil society and local populations, i s only beginning to take root. However, local governments are now required to engage local populations in the planning and monitoring o f local development plans (see chapter 4). Ongoing training by community drivendevelopment projects as well as an evaluation o f local accountability mechanisms (currently planned) should help to clarify how these mechanisms can be strengthenedin fbture. 13 Controlover deconcentratedline agencies 2.23 Control over deconcentrated line agencies is mostly limited to vertical financial control. This preponderance o f vertical control mechanisms is typically present in African countries (see Ndegwa 2003), where central government structures provide most public services. Within this vertical control structure, the ContrGZe des Dkpenses Engagkes (CDE) and the Commission CentraZe des March& (CCM) provide internal control, ensuring that spending commitments conform to budget allocations and procurements are carried out correctly. The IGE i s the public auditing body and has a very broadmandate that also includes public enterprises. 2.24 Inpractice, most of these control functions are constrained by cumbersome procedures and a general focus on ex ante (at the stage o f budget commitments) rather than on ex post controls. As a consequence, there i s persistent evidence o f leakage o f administrative funds in deconcentrated line agencies; the leakages are primarily caused by procurements at above-market prices and the failure to distribute goods to local service facilities. This is particularly true for the district administrations o f the health and education sector, where most deconcentrated funds are managed.l3 REGIONAL PLANNINGBODIES 2.25 Inmany countriesthat have decentralized to the local level, there is a tendency to regroup local functions at the regional level. Insome cases, regional bodies were formed to lobby for and coordinate the interest o f local governments and civil society (Ecuador, Bolivia, Peru). Inother countries, regional planningbodies help to improve sub-national monitoring and evaluation as well as planning and coordination among local governments and central administrations (i.e. Rumania, Bulgaria, Venezuela, and Colombia untilend o f 2002).14 2.26 Several regional and supra-communal initiatives have emerged inMadagascar in recent years: (i)Regional associations were set-up in the regions o f Ft. Dauphin, Morondava, and Moramanga. These associations identify regional development priorities and coordinate public investment as well as community development initiatives. (ii)Provincial planning committees in Fianarantsoa and Majunga have brought together statistical units o f deconcentrated service providers, civil society, and provincial authorities in order to coordinate data collection and information management. (iii) Inter-communal associations (OPCI) were formed by the semi-urban communes around the capital Antananatrivo, the communes in the district o f Ambato Boeny, and the communes along the railway Fianarantsoa-Manakara(see Box 2.1.). The OPCIs coordinate communal development plans, carry out joint investments and l3 thisrespectMadagascarfollowsthefrancophonecontrolsystem. Formoredetailsontheprocessof In budget execution see chapter 3. On the leakage o f public funds, see INSTATMADIO, forthcoming studies. l4 Frank2002forexperiences fromLatinAmerica. Cf. 14 activities, and represent local governments vis-&vis national govemment entities and donors. Box 2.1:The Inter-communal association Fianarantsoa-Manakara The OPCI Fianarantsoa-Manakara regroups all communes along the railroad between these two cities. For many communes the railway line represents the only regular form of access, ensuring the export of agricultural products fi-om the communes and the mobility o f the population. The main objective of the OPCI i s therefore to coordinate the protection of the track. To this end, OPCI organizes communes and communities to carry out maintenance functions and reforestation inerosion endangered areas along the track. Inaddition, the OPCI has formed a joint institutional capacity-building unit, which helps the participating communes to improve their administrative and financial management. Like all other intercommunal associations in Madagascar, this OPCI was created with external donor support (the U S NGO PACT) and still receives support from this side. 2.27 Typically, bottom-up processes drive these initiatives. They have no clear legal status and very limited own revenue sour~es.'~Yet, particularly in the case o f regional M&E and planningunits,this "lighter" institutional set up has facilitated the cooperation o f actors from the sphere o f decentralized and deconcentrated administrations that might not have collaborated under a more established institutional structures (i.e. municipal govemments, deconcentrated public sector providers, business associations, etc). 2.28 Inthe current Malagasy institutional environment o fdecentralization, OPCIs and regional bodies can be considered as a viable alternative to full-fledged intermediate governments. However, their replicability in other areas is subject to a number of constraints: 0 Financing. Giventheir informal character, most regional planning units do not have their own sources o f income. As a consequence, most o f the projects inMadagascar have remained dependent on external sources o f finance. Insome cases this has put at risk the continuity o f administrative services and made OPCIs and planningunits vulnerable to external factors such as changing donor objectives and stakeholder interests. A solution to this problem has been found by a number of OPCI that have begunto finance administrative overheads through regular contributions o f member local govemments. 0 Diversity across regions. Internationalexperience shows that sub-national planning and coordination mechanisms cannot be easily replicated or scaled-up through central govemment decree. Inter-communal associations in Madagascar for instance are often organized around very specific objectives such as the protection l5 thestatusofOPCIisprotectedbypresidentialdecree(Decret99-952). Only l6 Bottom-up driven municipal associations, in Ecuador for example, have been highly successful. This success has even fueled demand for more decentralization o f functions to intermediate governments. In contrast, centrally driven regional planning agencies in Colombia were less effective and had to be abandoned (cf. Frank 2002). 15 of the southern railway line in the case o f the OPCI Fianarantsoa-Manakara, or a joined investments plan inthe case o f the OPCI Ambato-Boeny. 0 Legal uncertainty. Due to their unclear legal status, regional planning units are particularly vulnerable to changes in the existing legal framework. Therefore, the success o f these projects inthe future will depend strongly on a clarification o f their legal status by national law. TOWARDS A MORE SERVICE-ORIENTEDADMINISTRATION 2.29 Given the constraints o f the Malagasy decentralization environment, this report recommends (i) a series o f short-term reforms intendedto make the existing set-up more efficient before decentralizing further, and (ii) and institutional reforms that should legal help to resolve existinginconsistencies inthe constitutional and institutional framework inthe mediumto longterm. Short- term reforms: Make the existing set-up more efficient before decentralizing further 2.30 In the short term, the Government should consider delaying decentralization to an intermediary level o f government, at least until a more reliable legal and institutional framework can be put in place. The immediate focus should be to strengthen the communes and to increase the effectiveness o f the existing system o f deconcentrated service delivery. 2.31 Strengthen the status of communes and encourage inter-communal associations. The government should review the existing legislation on local governments in order to clarify the rights and responsibilities o f communes. Particular attention should be paid to the competencies o f local governments, financial management and procurement, as well as revenue assignments and local tax collection mechanisms (see chapter 4). In addition, the status of inter-communal associations should be strengthened and communes encouraged to form new OPCIs where needed. 2.32 Strengthen the control and monitoring system for decentralized governments. Control over local governments is virtually non-existent and would have been potentially weakened further by the reforms o f the 1998 constitution. Inthe short term, the government should review the legal control functions o f local state representatives (sous prefets) and strengthen the mandates o f local treasury officials and representatives o f the Ministryo f Decentralization who carry out financial control over local governments. At the same time, training and technical support should be provided to horizontal control institutions such as commune councils and community groups. 2.33 Improve coordination at the central level. Even after the creation o f a self- standing Ministryo f Decentralization, the government needs to clarify the competencies and reporting relationships within the central administration with regards to decentralization policy. A possible division of labor could consist o f entrusting the Ministry of Decentralization with the planning and implementation o f legal, fiscal, and administrative decentralization policy, while the Ministry o f Interior would carry out 16 reforms o f the deconcentrated control fiamework. Reforms o f local tax collection mechanisms should be designed in coordination by the Ministry o f Decentralization and the Ministry o fFinance. 2.34 Improvetransparency and accountabilityof deconcentratedline agencies.In the short term, the government should focus on reducing administrative inefficiencies and leakages at the district level. Budget tracking surveys planned for the forthcoming PERand health sector work will provide important insightson the extent to which public funds reach the facility level. Inthe meantime, it is recommendedthat budgets allocated to school and health facilities within districts be publicly posted at the facility and commune levels. At the district level, budget information should also be made available to localbeneficiaries through participatory planning andmonitoring exercises.l7 2.35 Strengthen M&E and planning capacities within line agencies. In order to include local information in budget planning and programming, it i s necessary to strengthen sub-national monitoring and evaluation capacities. The most important reforms would be to establish clear reporting rights between statistical services at the province and central levels and to harmonize data collection procedures across administrative levels and sectors. In addition, investments are needed to allow for adequate data analysis, treatment and verification at the province and district levels, and to allow for better coordination mechanisms for information sharing with central statistics departments andbetween sectors. Medium- to long-term reforms: Revise the constitution and limit the number of sub- national layers 2.36 Inthe medium to longterm, the government should engage inbroaderreforms of the legal and constitutional framework for decentralization. In order to appease and adequately address conflicts that were at the root o f the crisis o f 2002, these reforms should be embedded ina broad-based andtransparent decision making process. 2.37 Clarify the principleof sub-national autonomy in the constitution. Giventhe experience o f the recent political crisis, a clearer definition o f the concept o f sub- national autonomy should be sought. In particular, the future status o f intermediate governments should be revised, as well as the distribution o f functions and legislative competencies between national and sub-national governments and the organization o f intergovernmental relations (central government-intermediate governments-communes). 2.38 Limit the number of intermediate governments to one. Each new layer o f decentralization automatically increases overheads and coordination costs o f service delivery. Given that the Malagasy government has very scarce resources to distribute, it l7Reforms o f the vertical control mechanisms will be discussed in the Country Financial Accountability Assessment (CFAA). An integrated financial management system will be implemented in a forthcoming public sector reformproject financed by the World Bank (Institutional Development Technical Assistance Project). Inthe long term, strengthened control, audit, and budget management systems will constitute an important pillar for decentralizing institutional responsibilities further. 17 i s advisable to consider the creation o f only one intermediate layer o f government at either the province or regional level. 2.39 Define a homogenous framework for the management and control of sub- national governments. While each sub-national authority should inprinciple report to the next higher level o fgovemment, there is a strong needto define a homogenouslegal framework for the distribution o f expenditure, resource assignments, and the control o f sub-national governments. A coherent national legal framework is particularly needed at the commune level, given the strong legal uncertainties associated with the previous transfer o f control and oversight rights to provinces. 18 3. FISCALFRAMEWORK THEFISCAL GAP 3.1 At the beginning o f the 21Stcentury, developing and transitional economies are fiscally more centralized than industrial countries were in their early stages o f development. In most developing countries, the central government receives more than 90percent o f total revenues and directly accounts for over 70 percent o f total government expenditures. l8 While expenditure responsibilities can be decentralized readily, revenue-raising capacity tends to remain a central prerogative. With regard to sub-national levels of government, this translates into a major challenge o f decentralization - the fiscal gap, or the mismatch between increasing expenditure responsibilities and insufficient generation o f own revenues. 3.2 In Afhca, the challenges of fiscal decentralization are further compounded because the overall revenue base is very low, which limits the transferable pool o f resources, and because most broad-based and productive revenue sources are also hardest to decentralize: 0 Low revenue base: On average, government revenues hardly reached 26 percent of GDP for SSA in 2001; in African HIPC-countries the revenue-to-GDP ratio was as low as 18 percent in2001;" 0 Centralization of revenues: A few revenue sources, such as trade taxes, the VAT, and personal income taxes, typically account for the larger part o f total revenues. For macroeconomic, equity, and efficiency reasons, it is widely considered that these revenues should not be decentralized. 3.3 In Madagascar, both indicators are below the - already very low - African average (see Chart 3.1 and 3.2): 0 Low revenue base: Government revenues represented only 9.6 percent o f GDP in 2001 - a figure which puts Madagascar among the lowest performers in Africa. During the 1990s, revenue-to-GDP ratios remained below 10 percent and only started to pick-up modestly since 2000 (see Chart 3.4); 0 Centralization of revenues: A few taxes account for the largest share o f total revenues. In 2000, the VAT, the Corporate Income tax and foreign trade taxes Anwar Shah (2001), "Fiscal Decentralization in Developing and Transition Economies: Progress, Problems and the Promise". Paper presented at the University o f St. Gallen. l9Ifnototherwisestatedalldatareflects"Government revenuesexcludingallgrantsasashareofGDP" and is taken from: SIMA Africa Regional Database. 19 accounted for 78 percent o f total tax revenues. Foreign trade taxes alone accounted for some 52 percent o f tax revenues (even after the elimination of all export duties in 1997). This share, which i s among the highest inAfrica, results ina large degree of centralization, as trade-tax revenues are typically considered to be indivisible. Consequently, the greater the reliance on trade taxes, the greater the fiscal gap is likelyto bebetweensub-national revenue andneeds. Chart 3.1:Madagascar's revenues inthe international context Government revenues as a share of GDP (excluding grants) 1998 30 20 10 0 Source: World Bank Central WBI Database Chart3.2: Madagascar's revenuesinthe international context Government revenues as a YOof GDP (excluding grants) 2001 45 40 35 30 25 20 15 10 5 0 Source: World Bank Regional Database 3.4 The centralization o f revenues i s mirrored by a very centralized system o f public expenditures, although some steps were taken to deconcentrate the execution of 20 expenditures in the late 1990s. In total, the central government manages and executes some 87 percent o f the total government expenditures, and 81 percent o f the recurrent expenditures. Deconcentrated expenditures amount to 10 percent (recurrent: 16 percent); they are executed by deconcentrated line agencies o f the ministries, mostly inthe social sectors. Investments are almost exclusively planned and executed at the central level. Decentralized expenditure amounts to 4-5 percent o f the total, out o f which 3 percent represent transfers o f the central government to sub-national governments and only 1-2 percent are sub-national own revenues (see Table 3.2). 3.5 The large contribution o f transfers to the sub-national governments' expenditures, is an indicator o f the magnitude o f the existing fiscal imbalance - an imbalance which i s even more strikingwhen viewed in the light o f the very low degree o f expenditures decentralization in Madagascar. If the government decides to progressively decentralize larger expenditure items to sub-national govemments (as was already envisaged for the autonomous provinces in2001), the country will be facing new and even greater fiscal challenges. Inthis case, improving the revenue raising capacity o f sub-national levels and developing an intergovernmental transfer system altogether transparent, predictable, and conducive to enforcing hard-budget constraints would be a highpriority. EXPENDITURES GeneralPrinciplesof ExpenditureDecentralizationandAssignment I nformation advantage and allocation efficiency 3.6 The main argument for the decentralizing expenditures is based on the concept o f information advantage and allocative eficiency. 2o Local agents can often identify more efficient, cheaper, and more appropriate ways o f providingpublic services. Although the central government may reap economies o f scale in collecting information, in practice democratically elected local governments still retain the informational advantage because o f political accountability- local politicians have a greater incentive to use local information (than national or provincial politicians have) because they are potentially directly answerable to the local electorate. 3.7 However, local govemments are answerable to the local electorate only insofar as the appropriate checks and balances are in place. As described in Chapter 2, local governments may be better informed about the preferences o f the local population, but they may not be sufficiently accountable to their local constituencies. Moreover, they may not have the capacity to handle the transfer o f responsibilities without support from higher levels o f government. Thus the information asymmetry exists at two different 2o See Wallace E. Oates (1972), Fiscal Federalism; Wallace E. Oates (1999), "An Essay on Fiscal Federalism", Journal of Economic Literature, 37, p. 1120-1149; Richard Bird (2000), Intergovernmental Fiscal Relations in Latin America: Policy Design and Policy Outcomes, Inter-American Development Bank; World Bank 1999, World Development Report 2000, Entering the 21" Century, Chapter 5: "Decentralization: Rethinking Government", Pranab Bardhan (2002a) "Decentralization and Governance", Journal of Economic Perspectives vol. XX; see also Bardhan 2002b. 21 levels - while the central government may not know what do to, the local government may not know how to do it.21 Theprinciple of subsidiarity 3.8 The "principle o f subsidiarity" states that expenditure decisions should occur at the lowest level o f government that can fully capture the costs andbenefits o f providing a public service. According to this principle, local governments should be in charge o f managing and providing local public goods. A contrario, higher levels o f governments should provide those services that entail economies o f scale, or that intemalize externalities across jurisdictions, and carry out those expenditures that involve clear stabilization, distribution, or standard-setting objectives. Ihousing Air andwater pollution Policeprotection Source:H.Fuhr; WorldBank Decentralizationhomepage, http://www1.worldbank.org/publicsector/decentralizatiod 3.9 It is typically considered that central government should retain control over at least national public services, international affairs, monetary policy, general regulation, transfers to persons and businesses, fiscal policy coordination, regional equity, redistribution, and the preservation o f internal common markets. Intermediary levels o f government may play an important role in providing some important social services (such as education and health), in inter-communal issues, or in providing oversight o f local governments. Finally, lower levels o f government should be responsible for all -or most - local services. Within the same field, different functions may be assumed by different levels o f government. For instance, in education: central government may retain authority on general policy and oversight issues; intermediary levels may take 21See among others Bird, Richard(1995); Crook, R.C., Manor J. (1998) Democracy and Decentralisation in South Asia and West Africa: Participation, Accountability, and Performance, Cambridge University Press; Bardhan, P. and D. Mookherjee (2000) "Relative Capture of Governance at Local and National Levels", American Economic Review, vol. 90(2). 22 over personnel-related decision while local governments could be made responsible for infrastructure provision and maintenance. 3.10 Whichever the chosen degree of expenditure decentralization, it i s vital that respective expenditure responsibilities be defined with great clarity. Failure to do so would likely result in duplications and overlaps as well as in detrimental bargaining between various levels o f government. In addition it i s important to apply the `)%"e follows functions " principle, which states that expenditure responsibilities need to be defined before revenue-raising authority i s distributed at each level. Typically though, the reverse is true - revenues are distributed with no, or little, concem for desirable distribution of expenditure responsibilities. The result is a weak budget constraint, a fiscally overburdened governments, andwidespread misallocation o f scarce resources. Madagascar'sSub-nationalExpenditures 3.11 Decentralization o f administrative and expenditures responsibilities are closely linked. As a consequence, the analysis o f Madagascar's expenditures set-up is complicated by the uncertainties surrounding the future o f Madagascar's decentralization policies. Currently, very limited decision-making powers have been transferred to local governments and a clear strategy for the transfer of specific hnctions to sub-national governments has not been outlined yet. In this uncertain setting, the purpose o f this section is to (i)provide an overview o f the current extent o f expenditure deconcentration, (ii) evaluate budget execution performance, and (iii) potential discuss expenditure reallocation within the existing framework. Trends and composition 3.12 Madagascar's expenditures are heavily centralized. Given the very weak revenue mobilization o f communes, the central government budget accounts for more than 98 percent o f total expenditures. These central government expenditures can be broken down into (i)centralized expenditures; (ii)deconcentrated expenditures, and (iii) decentralized expenditures (see Table 3.2): (9 Centralized expenditures account for 86.6 percent o f total expenditures. Investments are almost completely centralized; the transfer to the districts (representing 1.8 percent) has been suppressed since the beginning o f 2003. (ii)Deconcentrated expenditures account for 10.0 percent, mostly for recurrent expenditures in health and education. These expenditures will be specifically analyzed inthis chapter; (iii)Decentralized expenditures account for 3.2 percent of the total and represent the transfers to communes. Decentralized expenditures will be analyzed inchapter 4. 23 3,000,514 95.1% 82,679 2.6% 5,955 9,453 47,070** 1.5% TOTAL 129,749 4.1% 15,658,610 100% I 3,351,945 100% I 2,306,665 100% I Note: Datais from the Ministry of Budget's Tableaude Bord for the year 2001. Figuresare inmillion FMG. Transfers are administeredby the Min.ofBudgetand defined as centralexpenditures inthe budget nomenclature. !n the present table they are consideredseparately **Transfersto provincesandcommunes(healthand education) 88,990, transfers to communes(recurrent) 97,550 Transfers to districts (Fivondronana) 47,070 3.13 Deconcentrated recurrent expenditure has increased since the mid-l990s, when sector ministries increased their outreach. The degree o f deconcentration varies across ministries. In 2001, only three ministries had a "deconcentration ratio" of more than 20 percent: The Ministry o f Education (70 percent), the Ministry o fHealth (42 percent), and the Ministry of Transport (32 percent). All other ministries remained highly centralized, most of them with more than 90 percent of centralized expenditures (see Chart 3.3). Chart3.3: Shares of recurrent expenditures executed below the center 80.00% P Ministry Source: Bank staff calculations based on the allocations inthe Tableau de Bord for the year 2001. 24 3.14 The analysis below focuses on the composition o f deconcentrated expenditures in health and education, which account for 78 percent o f total deconcentrated spending. The relatively high shares o f deconcentrated expenditures inthese two ministries are due to the allocation o f salary payment authority to the provinces and allocation o f non- salary recurrent expenditures to the districts: 3.15 Education is the most deconcentrated sector with 70 percent o f deconcentrated recurrent expenditures. Most o f these recurrent expenditures (80 percent) are salary expenditures, out o f which only the administrative salaries (30 percent) are managed by the central ministry. Provinces and districts manage the remainder o f recurrent expenditures. 0 The province level administration (DIRESEB) administers 70 percent o f teacher salaries, which represent 55 percent o f total recurrent expenditures; 0 The district level administration (CISCO) administers 14 percent o f recurrent expenditures. Districts receive more than halfo f the ministry's non-salary recurrent expenditures. Two thirds o f their budget i s allocated to the purchase o f services (mainly current maintenance), while the purchase o f goods is allocated in equal parts to administrative and school supplies respectively. The center sets the share o f the budget that should get to the facilities inthe district (on average 30 percent to primary schools for school supplies, and the remaining to districts for recurrent maintenance). Therefore, in principle districts have some discretion on how to allocate the funds for a subcomponent o f their budget (Le. from maintenance to local teachers' salaries), but they should be conforming to the central allocations to facilities when distributing school supplies. 3.16 Health is the second most deconcentrated ministry with 42 percent o f recurrent expenditures deconcentrated. Like in the education sector, administrative salaries are managed at the central level, while the salaries o f health personnel are dealt with at the province level. Health districts manage parts o fthe non-salary recurrent expenditures: Chart 3.4: Socialsector spending by administrative level 31% EAlEtim Source: Bank staff calculations based on the allocations inthe Tableaude Bord for the year 2001: recurrent expenditure data 25 0 The province level administration (DIRISAN) is assigned a large share o f salary expenditures amounting to 30 percent o f total recurrent expenditures; 0 The district level administration (SSD) administers 12 percent o f total expenditures. SSDs receive 25 percent o f the non-salary recurrent expenditures. As in the education sector, most of the districts' budget (two thirds) is earmarked to the purchase o f services, mainly for current maintenance. Administrative supplies and expenditures for fuel account for most o f the purchase o f goods, leaving a small fraction for health supplies. Geographical allocation and Budget execution 3.17 The subsequent analysis focuses on two critical measures o f service delivery performance: (i) the equity o f geographical allocations (inthe education sector), and (ii) the efficiency o f budget execution to the health and education facilities. Both areas will receive additional attention inthe upcoming PER and a Health Sector Study. 3.18 Equity of geographical allocations. Existing studies show that there is a strong geographical dimension o f poverty in Madagascar, with communes and regions in the same province exhibiting huge variations in standards o f livings.22This implies that different jurisdictions will have different needs o f public service delivery as well as different abilities to provide services and to raise own revenues to finance their expenditures. 3.19 In order to measure the equity dimensions of the distribution of allocations, recurrent deconcentrated expenditures have been matched with indicators o f welfare across different geographic administrative units. To this end, a spatial description o f the distribution o f poverty was carried-out in Madagascar by combining the 1993 sample survey data (Enqugte Permanente des Me`nages) and 1993 census unit records. On the expenditure side, given that only recurrent non-salary expenditures are deconcentrated down to the district level for the social sectors, one can rely on complementary administrative data on personnel allocations to obtain an overall picture o f the total recurrent budget allocated to different geographical areas. (i) Thegeographic allocation of teachers (and their corresponding wage expenditures) is regressive: 0 At the primary school level, Madagascar has an extensive coverage o f schools with a disproportionately high share o f small schools in isolated rural areas. However, the allocation o f teachers does not match this high density o f public primary schools, resulting in a strong urban bias in teacher deployment: poor rural areas 22 "Putting Welfare on the Map in Madagascar", J. Mistiaen, B. Ozler, T. Razafimanantena, J. Razafindravonona (Africa Region Working Paper Series, No.34, The World Bank, August 2002) The fact that the poverty data dates back to 1993 calls for some caution in the interpretation o f the results. However, despite the fact that the census and household are from 1993, the poverty mapping exercise i s still meaningful. While the poverty levels might have changed over time, evidence from the 2001 commune census suggests that the relative geographical positionof poverty incidence has been stable in Madagascar over the past decade. 26 have fewer teachers per school and more pupils per teacher. These imbalances implya systematic lower quality o fthe service providedinrural areas. 0 At the secondary school level, public schools are also evenly distributed across the country although private facilities tend to be located inricher areas. As inthe case of primary education, there is an evident misallocation o f teachers towards districts that are richer and districts with a higher share o f urban population. Per capita allocations to non-salary expenditures in secondary education are reinforcing differences in the allocation o f teachers, reaching both relatively richer and more urbanized districts. (ii) The geographic allocation of non-salay expenditures is also regressive. Non- salary expenditures are distributed proportionately to the population across districts. Given the high variability in their poverty incidence across districts a uniform allocation to districts results inregressive expenditures. 3.20 These results are in line with existing evidence from the Education Sector which highlightsthe potential for important economies o f scale inbothprimary and secondary education. However, given the sparsely dense population inMadagascar, a large fraction o f public primary schools in rural areas are too small to be able to take advantage o f their potential economies o f scale and have unit costs o f service provision that still very high. 3.21 Inthe mediumterm, ongoing reforms that encourage teacher redeployment inthe education sector (multi-grade teaching and school consolidation) might allow schools to lower their costs. In the meantime, the geographical disparities call for an improved allocation towards poorer and rural areas that have a larger need o f unmet needs, both in terms o f teachers' time and availability, as well as in terms o f other inputs, such as instructional material. 3.22 Efficiency of budget execution. Inthe current deconcentrated structure, districts receive a small fraction o f the budget to provide materials to and basic maintenance o f health and education facilities. As examples from other countries have shown, districts may not have the incentives to actually distribute their budget to the facilities and be heldaccountable for it.24A budgettracking survey which was carried out bythe MAD10 23The Education Sector Study demonstrated that an important determinant o f the high drop out rate is supply constraints - the shortage o f teachers is such that many schools do not offer a complete cycle o f instruction. Accessibility and low quality o f service delivery are reflected by an entry rate at grade 1o f 81 percent and a survival rate at grade 5 o f 33 percent in 1997. See inparticular: "Education and Training in Madagascar. Towards a Policy Agenda for Economic Growth and Poverty Reduction", Report No. 22389- MAG. 24Reinnika and Svensson (2001): "Explaining Leakage o f Public Funds" (World Bank Policy Research Working Paper) show that on average during 1991-95 schools received only 13 percent o f the central government's allocation for the schools' non-wage expenditures from the districts. 27 project, gives some preliminary indications o f how the budget flows from districts to the facility 3.23 Inthe education sector, schools receive deconcentrated funds through the district (CISCO) for maintenance as well as a direct allocation for schools supplies. The MAD10 survey found substantial leakages for both categories, and particularly substantial for the CISCO allocation. 3.24 Inthe healthsector, the percentage ofexpenditures reachingthe healthfacilities (Centre Sanitaire de Base, CSB) was found to be higher than in the education sector. However, there was a substantial geographical variability: in some areas, the districts distributed supplies to the health facilities only on demand from the facilities themselves. Because o f poor communication and transport facilities, the poorer and more remote facilities tended to receive the fewest supplies from their respective district. 3.25 In both sectors, geographical isolation is among the main constraints to an effective execution of the budget. Inthe education sector, schools with a higher pupil- teacher ratio and more remote schools are less likely to receive the allocated budgets. However, the differences are particularly accentuated within rural communes and to a lesser degree between rural and urbanareas. The health sector presents the same features -thefractionofthebudgetreceivedislowerforbothruralandformoreisolatedareas. 3.26 L o w execution rates inremote facilities are explained by the long distances each budget manager (gestionnaire) needs to travel in order to go to the Circonscriptions FinanciBres (CF)26 to execute his budget. In theory, the budget manager for those districts that are far from the CF can use a simplified procedure to expedite budget execution (ddkgation de crkdit). Yet in practice, this simplified procedure is seldom used and only for a very small fraction o f the total non-salary credit o f the social sectors (3 percent ineducation and 0.5 percent inhealth for the 2000 budget).27 REVENUES Challengesin sub-nationalrevenuemobilization Efficiency versusAccountability 3.27 In most poor developing countries, two principles o f fiscal decentralization collide. While fiscal accountability calls for a close match between revenue means and revenue needs, the predicaments o f fiscal federalism revenue assignment amount, in 25The survey was fiianced by the European Commission and implemented by the national statistical institute INSTAT. The survey was carried out in selected districts in 2000 and 2001 for the budget years 1999and 2000 respectively (Madio Rapport MCthodologique, July 2001). 26The contr6leurs des dkpenses engage`es (for commitment), the sous-ordonnateurs (for ordonnancement) and the tre`soreries provinciales (for payment) are all based at the Circonscriptions Financikres (CF). There are 13 CF inMadagascar, 11o f which are located inthe provinces "Figures from "Madagascar: ExecutionBudgCtaire -Le tableau debord", byIanLienert andMarc Paoletti, IMF, Fiscal Affairs Department, September 2001, p. 49. The de`lkgation de credit i s mostly used intheMinistdredeForcesArmke andbythe MinistBre desAfaires EtrangBres. 28 practice, to allocating the collection and administration o fmost major taxes to the central government and to leaving local governments with only minor tax items: Matching revenue means as closely as possible to revenue needs. From a fiscal accountability perspective, financing should occur where the services are being provided. This could either happen (i) directly through cost-recovery, or (ii) indirectly through local taxation. 0 Optimal revenue assignment: Central government should raise taxes that (i) are mobile (like capital); (ii) redistribute wealth (like progressive taxes on income and wealth); (iii) are based on unequally distributeddeposits o f natural resources (like oil, gas, or minerals); and/or (iv) are critical for macroeconomic stabilization. Customs duties are also typically considered to be national-level taxes because o f their indivisible nature and because o f their strategic importance to foreign trade and national industrial development. Sub-national governments, as a consequence, must rely essentially on taxes on immobile factors (land and real estate), user charges, and tolls, most o f which yield only small or insignificant amounts (see Table 3.3). Central-level State-level Local-level Trade taxes Surcharges on national taxes User charges Value added tax Retail sales taxes Licenses and fees Corporate income tax Property taxes Naturalresources taxes Vehcle taxes 3.28 Inmost developing countries, this "optimal" assignment o f taxes results directly inthe observed over-centralization of revenues. The central government level controls high-yield taxes that account for the most part o f total fiscal revenue. Because many expenditure items can be decentralized readily while revenue items cannot, the imbalance i s automatic. This gap is compounded in cases where decentralization efforts entail a transfer o f services responsibility to local governments in sectors with major expenditure items such as education or health. 3.29 The low revenue base o f sub-national governments and the resulting detachment o f means and needs has the potential to jeopardize the accountability purpose o f decentralization and carries the risk o f structural deficits at the sub-national level, potentially leading to macro-disequilibria (especially if sub-national units expect to be bailed-out). In turn, it threatens to turn on its head the entire rationale for decentralization, the core purpose o f which i s to allow for better and more efficient service delivery and for more accountability. 29 Bridging thegap 3.30 Sub-national governments need to increase their own revenues to bridge the gap and to overcome the accountability challenge. In order to increase own revenues, sub- national governments can adopt a strategy that focuses on (i) increasing the rate of existing taxes, (ii) widening the tax base, and (iii) improving the administration o f the existing mix. 3.31 Increasing tax rates may be appropriate in cases where taxes are revenue- inelastic (such as user charges on water or the property tax). However, as a strategy, it may be the least successful where low revenues are not explained by low rates o f taxation. In Madagascar low tax revenues are mostly due to tax evasion and/or exemptions as well as to the important share o f the total economic activity that remains unregisteredandun-monetized. 3.32 A convenient, and often advocated formula is one that consists o f sub-national surcharges on some major national taxes ("piggy-backing"), such as the central personal income tax or the value-added tax. Intapping into these revenue sources, sub-national - typically state - govemments could balance their budget while retaining an important element o f public accountability. However, in poor developing countries like Madagascar, revenue sharing formulas (including "piggybacking") seem unpractical and even risky if they amount to multiplying and diluting responsibilities. The implementation o f revenue sharing mechanisms is bound to be hampered by a lack o f administrative capacity, efficiency, or oversight mechanisms, leading to leakages and unpredictability.28 3.33 Widening the tax base is an option particularly relevant in countries where important revenues and capital are attributable to the informal economy. InMadagascar, this idea spurred the creation o f the Impat Synthe`tique, which targeted individual enterprises in the formal and informal sector with a turnover o f less than 12 million FMG (or some US$2,000). Given that this tax has only been implemented in2002, the year o f the political crisis, it i s too early to draw conclusions on its impact; yet preliminary analysis suggests that yields may be too low to justifi additional administrative costs. InMadagascar 10 out o f more than 25 tax instruments generate 95 percent o f fiscal receipts. This means that numerous taxes are used to collect marginal revenues, and it suggests that improvements in revenue administration may be more useful inthe short term thanthe multiplication o ftax instruments. 3.34 Improving revenue administration i s the most promising strategy, but it also encounters many challenges as it often entails a concerted effort o f sub-national govemments to develop or update cadastres as well as the central administration to supervise and validate. For the state level, the most promising sources for sub-national revenue mobilization are retail sales taxes and excises; for the local level, user charges, licenses, fees, andproperty tax offer the largest potentials. '*Fora more in depth discussion of piggybacking options see Bird (2000) and Brosio (2000). For the challenges oftax sharinginVietnam see World Bank 1996. 30 3.35 At the state level, retail sales taxes have become a key revenue source in industrial countries. In the United States they represent a third o f total state revenues. However, in transition and developing economies, experience has been mixed. In countries with a large informal sector, where most vendors do not have a fixed business location, the tax is extremely difficult to administer. 3.36 At the local level, financing is evenmore challenging. Consequently, most local governments depend on transfers from and/or tax sharing with higher levels o f government. User charges and property taxes are typically considered to be the most promising and stable source o f revenue for the local level. However, inpoor developing countries, property taxation has not yet yielded the anticipated results. In most o f Africa, results have been di~appointing.~' 3.37 Experience from other countries shows that substantial increases in property tax revenues only materialize ifthey are planned in a comprehensive way. Real property tax reform i s a process o f financing local governments - not a series o f unrelated functions. It relies on information and support from many administrative departments and requires political commitment from all levels o f government. For instance, the valuation o f local property will not have an impact on property tax collection ifan oversight institution has not validated the tax registry and/or ifcollection mechanisms have not been clarified. 3.38 There i s a need - especially incountries with low local administrative capacity - to drastically simplify the property tax schemes and to apply them uniformly. Instead o f applying high-tech methods in valuating property, participatory assessment procedures can significantly improve yields and reduce administration costs. The most promising results can be achieved when property valuation is linked to a process o f formalization o f property ownership. The latter point seems especially relevant in the Malagasy case where, for instance, only about 10 percent o f agricultural landhas formal titles. 3.39 User fees are also underutilized, although they could substantially recover parts o f the costs o f local service delivery (see Box 3.1 on South Africa). User charges offer the advantages o f creating a direct connection between consumption and cost, and of greatly simplifying collection, due to the direct nature of the charge as well as the fact that non-payers can easily be excluded. In addition their economic value i s increased by resulting efficiency gains - by providing information on demand to public suppliers, the suppliers can evaluate ifthe services are valued at their marginal cost. 3.40 Because o f equity concerns, however, it is often argued that user charges are more appropriate for services that have some private components rather than for services that have "public good" qualities. Where demand i s price-inelastic (demand for water, for instance), the introduction o f a fee would not result in significant economic distortions but its effect would be clearly regressive in equity terms. In contrast, a fee i s progressive for goods with price-elastic demand (such as electricity). This type o f taxes *'On the unfulfilled promises of property tax collection see Bahl(2002); on the potential of user fees see Smoke (2001). 31 could have an important impact in communes with local markets, on which local governments could levy market rights and service charges. Box 3.1: MunicipalRevenues inSouth Africa (2001) Madagascar's revenue challenge Trends and composition 3.41 The Malagasy national budget depends on a few main sources o f own revenues, all o f which are considered to be central revenues by nature. Trade taxes made up almost half (48 percent) o f total budgetary revenues in 2001, after having been consistently above 50 percent in the previous years. Therefore, most fluctuations in total revenues since 1988 are explained by a fluctuation in revenues from trade taxes (see Chart 3.5). The VAT alone represented 67 percent o f all taxes on goods and services and some 17.5 32 percent o f total budgetary revenue. Finally, taxes on companies' incomes and profits accounted for about 10percent o f total budgetary revenue.30 Chart 3.5: Central GovernmentBudgetaryRevenue 1988-2000 -m- Taxes on Income & profits+Taxes on property +Taxes on dom. G&S -+i+ onforeigntrade Taxes __ Source: IMFRED2003 3.42 The centralization o f Madagascar's revenues is higher than in most other countries in the region for which data is available. InMadagascar, the center collects more than.98 percent o f total revenues, leaving communes with 1-2 percent o f total revenues. This centralization o f revenues results from the structural imbalance in the composition of the revenues, which can only be rebalanced through a gradual, medium- to-long-term process of increasing sub-national revenues. Revenue Assignment 3.43 Madagascar's revenue assignment i s in line with principle considerations o f efficiency and accountability. Revenue items with macroeconomic, equity, and efficiency impacts are assigned to the center. Although the 1998 blueprint included some fiscal decentralization to the provinces (see Annex 3)' the new government has backtracked. In the 2003 Budget Law, only a few revenue items, such as the imp& synthe'tique, remain candidates for decentralization (see Table 3.4). The 2002 Budget Law introduced a 100 percent allocation principle for each tax, assigning revenues to one level only, excluding piggybacking or tax sharing. The new government still needs to decide ifit will uphold this principle. 3.44 Furthermore, even for those revenues that accrue to sub-national levels, tax policy and collection generally remain centralized. The absence o f sub-national control over tax rates strongly curtails the decentralization logic, as sub-national jurisdictions 30Source: IMFRED January 2003. 33 have no flexibility in adjusting taxation to reflect the expressed preferences o f their respective constituencies. Table 3.4: Revenue assignmentsinMadagascar * Destinedto the ** Collectivitb decentralise`es ***Splitbetween the Region and the Commune. Collected by the center for decentralizedinvestment projects. The state of Madagascar's sub-national revenues 3.45 Analysis o f sub-national revenues in Madagascar is complicated by the strong discrepancy between revenue allocations as they are spelled out inthe law and reality as it is observed on the ground. In setting up the autonomous provinces, Malagasy legislators also transferred important revenue items to this new level o f government. Although the provinces have not been effectively put in place, the texts have not been significantly amended. 3.46 Province: At the provincial level, the main revenue item is the taxe professionelle (TP). TP yields have shown the largest increase in recent years, but their importance varies greatly according to the provinces - they account for up to 97 percent o f provincial revenues in Antanarivo, for 70 percent in Toasamina, but only for 31 percent in Antsiranana (AfDB 2001). While the L o i de Finance 2001 also allocated several other taxes to the province level (including business taxes, the "Vignette", the transaction tax, and licenses for the sale o f alcohol), the new government has backtracked on all o f them. It is also unclear how the Imp& SynthCtique will be allocated. 3.47 Commune Level: Commune level finances will be addressed in depth in the next chapter, andtherefore this sectionwill only point out broad trends and findings: 0 Total commune revenue i s minimal, representing 3-4 percent o f total government revenues; rural communes represent 1-2percent; 0 Madagascar i s facing a rural-urban divide - while many urban communes have increased their own revenues substantially since 1996, very few rural communes 34 have demonstrated a similar performance. In urban communes, own revenues represent approximately 75 percent o f total revenue, in rural communes only 25 percent; 0 The poorer the commune, the greater the reliance on economically more counterproductive and volatile revenue sources such as sales taxes on local primary goods. TRANSFERS Filling the sub-national gap 3.48 Transfers are the financial lifelines o f most sub-national governments. This prominence o f transfers in the mix o f local financing instruments results directly from the structural imbalance between sub-national own revenue mobilization capacity and expenditure assignment. Transfers are needed to bridge the sub-national fiscal gap and ensure that local and provincial governments can function and deliver an increasing range o fpublic services. 3.49 Across sub-Saharan Africa, transfers account for the overwhelming share o f sub- national revenues - while a few urban govements can contribute significantly to their cities' expenditures, most rural communes and provincial entities are unable to finance even a small fraction o f their expenditure needs out o f own sources. 3.50 Intheory, other fiscal tools such as tax sharing and sub-national borrowing are available to close the sub-national fiscal gap. However, as presented in the previous chapters, tax sharing and borrowing are very limited and often not desirable in most African countries. Thus, inpractice the design o f the intergovernmental transfer system becomes the single most important tool for allocating resources to the sub-national level. Objectives 3.5 1 Intergovernmental transfers are a necessity in any country with some minimum degree o f decentralization. The optimal design and breadth o f the transfer system depends on the degree o f both revenue and expenditure decentralization - the more fiscally decentralized the country, the higher the demands on the transfer system. In other words, transfers are essentially by-products o f the assignment of revenues and expenditures across different levels o f government. Yet if transfers are fundamentally reactive bridging instruments, the choice o f different transfer designs can have a significant impact. Transfers can sometimes provide important incentives for service delivery; at aminimumthey should avoid creating disincentives. 3.52 The choice o f transfer instrument and design can be geared toward several different aims. Four o f these objectives are particularly relevant in developing and transitional economies. Well designed transfer systems can (i) provide incentives for efficient revenue and expenditure management, (ii)foster greater equity across 35 jurisdictions, (iii) implementation o f certain minimumnation-wide standards, and ensure (iv) provide corrective measures to e~temalities:~~ Efficiency: Provide incentivesfor eficient revenue collection andfiscal discipline. Intergovemmental government transfers should be structured so as to encourage (and at least not discourage), (i) local revenue collection efforts; and (ii) spending discipline. Transfers need to preserve hard-budget constraints at the sub-national level. They should not serve to bailout those governments who fail to adequately collect local taxes or to spend revenues efficiently. Equity: Reduce regional fiscal disparities. Transfers should redistribute national resources so as to match revenue effort and service delivery outcomes, wherever one lives. This i s even more relevant as increasing fiscal decentralization tends to stress disparities across jurisdictions. High degrees o f inequality between sub- national jurisdictions create a strong demand for progressive transfer systems targetingpoorer areas. Minimum Standards: ensure the delivery of minimum standards of key public services. The transfer system should help to ensure that minimum standards for public services will be achieved through sub-national financing. This applies in particular to basic services in health, education, water and sanitation, as well as to administrative services. Correction of spillovers. Transfer systems may be needed to support sub-national expenditure on items with strong spillover effects, for instance infrastructure. Without any incentives (e.g. matching), decentralized institutions are likely to minimize their spendingwith positive externalities. Design 3.53 The design o f a transfer system needs to reflect these broad objectives. At the same time, it also needs to take into account the country's level o f development. The poorer the country, the lower the administrative capacity (especially at the sub-national level), and the greater the need for simple, straightforward, and predictable transfer systems. 3.54 Designing transfer systems involves decision making in three areas. First, one needs to determine the transferablepool, or the overall amount to be transferred to sub- national levels. Once the overall envelope o f transferable resources has been defined, the next issue i s the distribution o f this transferable pool among sub-national governments. Finally, a decision needs to be taken regarding the type o f transfer instrument and the use o f conditionality. 3.55 The transferable pool can be determined (i) ad-hoc, for instance through an annual parliament-approved appropriation, or through a (ii) formula allocation based either on a simpleproportion o f total revenues or more complex formulas (including, for instance, an evaluation o f sub-national needs). Ifthe central government wishes to retain 31See Shah (1994); Bird (2000), Prud'homme (2003). 36 the greatest flexibility in its fiscal policy decisions and the ability to match transfers as closely as possible to actual needs, the incentive will be to choose the first option, determining the transferable pool annually and taking into account each year's new budgetary priorities. However, from the point o f view o f lower tiers o f government, "ad- hocism" is detrimental to proper sub-national budgeting and to maintaining a hard budget constraint -its discretionary nature leaves too much room for political bargaining and for the exercise o fvested-interest pressure. 3.56 The distribution o f transfers should (i) based on some clear and transparent be formula, and (ii) avoid excessive multiplication o f transfer programs addressing multiple objectives. A system o f ad-hoc allocations should be avoided as it leaves room for favoritism while underminingthe predictability o f the allocation process. 3.57 In poor countries with high degrees of inequality, it is desirable to use the transfer system as a tool to target the poor, while at the same time rewarding performance (inparticular the tax effort) o f the jurisdiction inquestion. The fundamental challenge o f such a system is that the twin objectives o f poverty alleviation and effort stimulation can become contradictory: the poorer jurisdictions, where needs are higher, typically tend to under-perform in terms o f revenue collection. Conversely, where revenue equalization is pursued with no regard for own-revenue generation effort, both collection efforts and prudent spending are likely to be discouraged, ending up in a situation where those who try the least get the most. One way out of this dilemma is to establish a measure offiscal effort that does not take into account actual tax revenue only, but tax revenues inrelation to local revenue generation capacity. 3.58 Yet even if it is possible to conceptualize a pro-poor, incentive-based transfer system in theory, several practical challenges may impede implementation. First, sophisticated formulas require abundant and reliable statistical information that is typically not available inpoor African countries. Second, the degree o f complexity o f the formula must take into account the level o f administrative capacity o f the country. Even in Latin America, where both statistical information and adequate capacity exist, measures o f fiscal effort have often proved to be too complex to measure and to be properly exploited. Third, targeting the poor through intergovernmental transfer systems has been difficult because poorerjurisdictions are typically worse at targeting their own poor (Ravallion 1999). 3.59 The use of conditionality i s widely debated. On one hand, unconditional grants to sub-national governments increase their flexibility and discretion to use the grants as part of their total budgetary resources. On the other hand, conditionality is often conceived as an insurance that funds will be spent for activities, that are considered a highnationalpriority, such as the provisiono fbasic social services. 3.60 In practice, effective use o f conditionality is challenged by the fungibility o f resources. Current experience from other Afkican countries suggests that the timing o f the earmarking along the decentralization process is crucially important. In Uganda an outright unconditional transfer o f responsibility for the delivery o f primary health care 37 resulted in a sharp deterioration o f spending in health care.32 In Tanzania, the initial earmarking o f grants has been gradually phased out towards more flexible forms o f transfers as local governments proved their management capacity and p e r f ~ r m a n c e ~ ~ . 3.61 The choice o f instrument depends on the degree o f decentralization and on the objectives of the transfer. As a general rule, the more responsibilities have been devolved to lower levels o f government, the higher the share o f conditional grant. For instance, if a country decides to devolve the management and payment o f teachers to lower level o f government, the corresponding transfer would be made conditional in order to assure adequate funding o f these functions (like salary payments). Table 3.5 provides a summary o f grant designs for the four core objectives defined above. Grant objective Featuresof grant design Maintaining incentives for efficient Generalpurpose matching grants revenue collection and fiscal Predictableand transparent formula-based discipline No bail out Reducing regional fiscal disparities Generalpurpose non-matching grants - Fiscal capacity equalization transfers (iffeasible) - Service delivery equalization (ifneeded) Achieving national minimum Specific purpose non-matching grants with conditions on standards standards o f service and access Compensating for benefit spillovers Specific purpose matching grants with matchingrate consistent with the estimated size o f the stillover Madagascar's transfer system Structure 3.62 Madagascar's transfer system was set up when the country embarked on the first phase o f its decentralization strategy in 1995. Since then, the 1392 communes and 111 districts have been receiving transfers. The transfer system was expanded in 2000 to include the newly created autonomous provinces. In 2001, these transfers amounted to 160billion FMGwhich represented some 4.7 percent o f total budgetary revenue: 0 The six provinces each received a subsidy o f 4.1 billion in 2001 (up from 1billion in 2000) for current expenditures during the transition until the full establishment o f the autonomous provinces (planned for 2002). A new transfer system, which took into account the transfer of expenditure responsibilities to provinces and also included horizontal equalization among provinces (through a solidarity fund), was also considered by the previous government. 32 See Akin, et a1(2001) "Decentralization and Government Provisiono f Public Goods: The Public Health Sector inUganda". http://www.cpc.unc.edu/measure/publications/workingpapers/wpO135.pdf 33 World Bank (2001), "The UnitedRepublic of Tanzania. Decentralization Sector Work", May 2001. 38 The 111 Districts each received a yearly transfer o f 406 million since 1995 to support local investments. In 2002, this transfer was programmed but not executed. A steering committee, the Comite'Local de De'veloppement (CLD), led by the deputy and the under-prefect, manages this subsidy. This arrangement is fairly atypical, as it does not target a decentralized administration but rather a core level o f the central administration. With the 2003 budget this subsidy has been phasedout andpartially re-allocated to communes. The 1,392 Communes receive four different transfers per year: one non- conditional block grant transfer to cover current expenditures, two smaller conditional grants for maintenance works inlocal schools and health centers, and one small conditional grant for the Secre'taire d'Etat Civil, who carries out administrative functions on behalf o f the central state. In 2002, the average transfer amounted to 64 million, which breaks down into 50 million (equivalent to US$7,500) for rural and 445 for urban communes (Table 3.6). Table 3.6: Transfers to administrative levels (inmillionMGF)in2001 Grant design Per unit TOTAL Province General purpose grant 4.100 24.600 District General purpose grant 406 45,066 Commune One general purpose grant 64 (average) 89,746 Three Specific purpose non- matching transfer Urban 445 (average) 25,000 (approx.) Rural 50 (average) 65,000 (approx.) TOTAL 159,412 3.63 The transfers to communes grew on average by 13 percent to account for population growth and inflation between 1998 and 2000, while transfers to the districts (Fivondronana) grew on average by 10.5 percent during the same period. Transfers to communes increased from 77.4 billion to 112.7 between 1998 and 2001. Transfers to the districts increased from 33.6 billion in 1995 to 40.3 billion in 2000 and 45 billion in 2001.34 3.64 The allocations to communes formula i s based and depends on the size and status (rural or urban) o f the commune. Every commune receives a minimum allocation o f about 34 million for the four transfers combined, irrespective o f size and status. All communes above 7,900 inhabitants receive a per-capita based allocation o f 3,800 FMG per inhabitant inrural communes and 8,800 FMGper inhabitant inurbancommunes (see table 3.7). 3.65 The unconditional subsidies from the central government (block grant) dominate the overall transfers that communes receive from the center (on average 81 percent o f the total transfers). Around 15 percent o f all transfers on average is accounted for by conditional transfers that are designated to support two of the communes' basic 34Source: Brunet 2003; Ministryo f Decentralization. 39 functions: administrative services and basic maintenance for health and education facilities. Table 3.7: Transfer to commune budgets(2001) Per capitaallocation Total (inmillionMGF) Grantfor current expenditures Urban:8,800; Rural:3,800 Urban: 19,862; Rural: (fonctionnement) (minimum:30million) 54,138 74,200 Grant EPP 402 4,872 Grant CSB 341 6,215 GrantSecrbaire de 1'Etat Civil 311,78 4,459 (minimum:2.5 million) Achievements 3.66 Most African governments are applying ad-hoc transfer systems characterizedby substantial discretion in allocation decisions by the center and frequent negotiations between the center and sub-national governments. In contrast, Madagascar's transfer system i s transparent, formula-based and simple, which makes the resulting allocations predictable and less subject to clientelistic bargaining. 3.67 The existing practice to calculate the allocation on fixed formula increased predictability even further. Madagascar's communes get an annual per-capita allocation, adjusted on a yearly basis for inflation and population growth. Such a system creates maximum predictability for communes with respect to expected revenues. There is no fiscal risk for the state as possible fluctuations in the annual share o f the transferable pool would not exceed 1 or 2 percent o f total revenues, equivalent to 0.1-0.2 percent o f GDP . 3.68 Infiscally centralizedcountries like Madagascar a formula-based transfer system i s clearly more desirable than an ad-hoc based system essentially for two reasons: (i) local governments crucially depend on transfers -therefore the predictability o f transfers amounts and availability i s paramount; (ii) the macroeconomic implication o f these transfers i s very low, which guarantees sufficient fiscal flexibility for the central government. Weaknessesand challenges 3.69 The main weakness with Madagascar's transfer system relates to actual execution (rather than design) o f the transfer system. These delays are often due to inefficient accounting arrangements between the Ministry o f Budget and local treasuries. The transfers typically arrive late inthe fiscal year, and sometimes not at all. Ina sample o f 15 rural communes, none has received the transfers on time in 1999 and 2000; in 1998, only two communes received their transfers as scheduled. Most transfers arrived 40 late by six months or more (see Chart 3.5). The delays for the conditional transfer (for the payment of the Secre`taired'Etat Civil) are evenmore important -a good thirdo fthe communes received their transfers even later than one year (years 1998-1999 only). These delays significantly reduce capacities for strategic planning and management in the majority of rural communes whose revenues depend most strongly on central transfers. 3.70 Madagascar's transfer architecture targets too many levels o f government, some o f which are not even decentralized. Until 2001, intergovernmental transfers were transferred to a decentralized level (commune), a deconcentrated level (district), and a level "in transition" (province). Under the 1998 strategy, a horizontal transfers system between the provinces (establishing a "solidarity fund") would have been set up, complicating the system even further. The transfers to the districts, which are a core unit o f central government service delivery, are particularly questionable. This transfer totals 45 billion FMG, which is equivalent to two thirds o f the total allocation to rural communes. However, this transfer has been used very inefficiently. Only a small share of the allocated 45 billion arrived at the local Chart3.6: Delayofunconditionaltransfers to rural communes (sample of 15 communes) Delays in Transfers to Rural Communes Sample of 15 Communes 1998-200 - 10 5 0 I1 1998 1999 2000 1 El No delay 1to 6 months 0 6 to 12 months EaMore than 12 months I Source: Miara Mita, Commune Finance Study 2002 3.71 The 1998 plan for inter-governmental relations created the risk o f actual re- centralization o f service del*ivery in Madagascar - the autonomous provinces were given the means and authority to exert extensive control and financial oversight over the communes. Provinces were to decide over the amount and mix o f communal revenues, with full discretion on the distribution o f sub-national revenue items and transfers. This system clearly threatened to deprive the communes of their financial lifeline. 3.72 While the allocation formula proved to be predictable, it is not governed by a permanent legal framework but depends on the annual budget law. This reduces the overall predictability o f the transfer system because transfer formulas are liable to ~ 35 See Chapter 3, table 4.2 and Miara Mita (2002). "Etude SUI les besoins financiers des Communes a Madagascar". 41 change every year. Inthe year o f political crisis (2002) for instance, total transfers were reducedto 21million per commune on an exceptional basis. 3.73 A last potential shortcoming relates to the formula o f allocating transfers to communes. Everything else being equal, the current per capita allocation formula is regressive as it penalizes rural communes. Even though rural per-capita income is substantially lower than in urban areas, rural per capita transfers are set at FMG 3,800 compared to FMG 8,800 for urban communes (see table 3.7). With 85 percent o f the population, rural communes receive 75 percent o f total transfers to communes (FMG 67,5 million) while urban communes, representing 15 percent o f the population, receive 25 percent (FMG 22,2 million). 3.74 There are several potential - and potentially conflicting - arguments for fixing the rates o f rural v i s - h i s urban transfers. Higher allocations for urban communes can be justified by efficiency concerns, while equity principles tend to militate for higher allocations to rural communes. While urban communes arguably face higher expenditure responsibilities, they also have access to substantially more revenues sources: 0 Revenues: Urban communes (i) larger own resources due to a higher have number o f business activities; (ii) benefit from more efficient tax collection on residential property; and (iii) benefit from local revenue collection o f the central government administrative apparatus (see also chapter 4); 0 Expenditures: Urban communes (i) expected to provide higher levels of are services - especially for the provision o f economic infrastructure; and (ii)demonstrate greater spendingefficiency. 3.75 In practice, the separate treatment o f rural and urban communes can be questioned. First, the existing separation between "urban" and "rural" communes i s artificial. For instance, many communes in the agglomeration o f Antananarivo are relatively urbanized, but classified as "rural", while smaller cities in the south are considered as "urban" although they tend to have predominant rural features. Second, the lack o f solid and timely empirical data, most notably on the respective unit costs o f service provision, impedes better targeting o f transfers between and within rural and urbancommunes. IMPROVING MADAGASCAR'SFRAMEWORK FISCAL Expenditures 3.76 Strengthen the monitoring and budget planning in social services. The responsiveness to local service delivery i s very weak. This suggests that there i s substantial room for improving the a l i m e n t between salary, non-salary, and investment expenditures to actual needs across districts. Correcting this mismatch can have substantial benefits for the poorest rural districts, where the mismatch i s the greatest. The unit o f analysis for planning allocations (both recurrent and investment) should be the schoolhealth facility. Information on local needs should be transferred to the line ministries before the budget is prepared. 42 3.77 Speed-up execution, commit budgets globally for a period of sljc months, and allow for reallocations between budget lines. Deconcentrated district administrations already enjoy some discretion in the management o f non-salary recurrent budgets. To further extend the outreach of central ministries, key services should be further deconcentrated with increased local decisionmakingautonomy inbudget management. 3.78 Improve the geographical incidence of current social expenditure in poorer areas. Current allocations o f deconcentrated expenditures need to be better targeted to poorer districts. This could be done by giving heavier weight to more remote and poor districts. As demonstrated by the poverty mapping exercise, districts are the optimal unit o f targeting as (i)they represent the core unit o f social sector service delivery and (ii) ratevariationsarestrongerbetweendistrictsthanwithindistricts. poverty Revenues 3.79 The key fiscal reform recommendations for the commune level are presented in the next chapter. Concerning the overall fiscal framework, the most important reform would consist o f confirming and actually implementing the 100 percent allocation principle for each tax (whereby each revenue item is attributed in full to one level of government). Experiences with tax sharing arrangements have not been positive in Madagascar. The expected revenue shares often did not arrive at the local level. In addition, tax sharing also breaks with the accountability principle - which also proved very relevant in Madagascar - that there should be the closest possible match between the collection and spendingunits. Transfers 3.80 Even though a more sophisticated transfer system, including poverty targeting, would be desirable, the first priority of the government should be to improve the mechanisms o f transfer execution. In a first phase, reforms should focus on providing relatively simple and practical solutions to transfer delivery problems. The impact o f those improvements on financing local services and on local administrative efficiency would already be substantial. 3.81 Inthis first phase, reforms shouldalso target the (i) simplification o f the overall transfer system; (ii) removal o fregressive distortions o f the current transfer system; and (iii)installation o f a monitoring system, including sanctions for non-compliance. 3.82 Inasecondphase (e.g.after three years), the transfer formula couldbereviewed inorder to include some measures of tax effort and poverty targeting into the allocation formula. However, attaining this level o f sophistication will remain dependent on the sine-qua-non condition that recent and reliable data i s available for all the 1,392 communes. Ifan intermediarylevel o f sub-national government were to be established, a pro-poor, incentive based system could be envisaged for this level first, as the necessary data would be easier to collect at a more aggregated level. 43 Priorities for the first phase i.Simplificationoftheoverallsystem 3.83 Transfers should go to decentralized levels of government only. Intergovernmental transfers are defined as transfers from the center to decentralized levels o f government. As long as provinces and districts remain deconcentrated outposts of the central administration, they should be financed as any other deconcentrated level o f government. The suppression o f transfers to districts (effective with the financial law 2003) should be madepermanent. 3.84 The center should keep responsibility and oversight over the transfer systems. The delegation o f transfer authority to the provinces (as in the 1998 blueprint) would have created major risks with respect to the overall objectives o f the transfer system (gap filling, equity) and it would have endangered the autonomy o fthe communes. 3.85 The mix between conditional and non-conditional transfers to communes is appropriate. So far, little expenditure responsibility has been devolved to the local level, so that there i s no need to increase the share o f conditionality o f the existing mix. However, when the 45 billion previously transferred to the districts will be reallocated, this increase should be made a conditional grant for investments. This transfer has already been earmarked for investments and would help communes to meet or exceed the legally required 15 percent investmentsthreshold (see chapter 4). ii.Improvepovertytargetingandtransparencyofthetransfersystem 3.86 Re-evaluate the transfer formula to rural and urban communes. Depending on the objective o f the transfer system (efficiency versus equity) different options for allocating transfers are possible. This paper argues for increasing the allocation to rural communes substantially, at least up to the per-capita allocations o f urban communes. The per capita transfer to rural communes would thus be increased from 3,800 to FMG 8,800 and the increase o f the minimum allocation to FMG 40 million be made permanent. This would also remove the artificial and regressive separationbetween rural and urban communes and help guarantee that rural communes have the minimum amount o fresources to function. 3.87 Finance this increase by reallocating the transfers previously assigned to the districts in favor of rural communes. Inthis way, the strengthening o f rural communes would be fiscally neutral while the transfer system would gain in simplicity and focus. Under the scenario o f a per capita allocation o f FMG 8,800 (and a minimum allocation of FMG 40 million), transfers to communes would increase by 43 billion to 133 billion (including the existing conditional transfers). This increase could be fully financed through the savings from the transfers to the districts. iii.Installationofamonitoringsystemincludingsanctionsfornon-compliance 3.88 Condition the transfers on effective reporting by the communes of both planned and executed budgets to the Ministy of Decentralization. Only a quarter o f communes have been regularly reporting their annual budgets and accounts to the center, thus contributing to and deepeningthe poor quality o f the monitoring system. 44 3.89 Use deconcentrated agents as supervisors and controller, not as executers of commune affairs. A more rigorous enforcement o f reporting duties would also imply a stronger control function for the deconcentrated agents o f the central authorities (in particular o f the under-prefect and the DAA).To efficiently carry out this function, these agents should be relieved from executing duties for which the communes have responsibility (such as the collection o f local taxes). Medium-term considerations 3.90 In the medium to long term, the general purpose component of the transfer system could be refined to include both poverty and performance indicators. This, however, is a difficult undertaking, which also runs the risk o f overburdening the transfer system with objectives that are beyond what the whole decentralization framework can achieve. The minimum condition for introducing such a system are (i) the availability o f reliable data for all jurisdictions (which depends on establishing a monitoring and evaluation system); and (ii) the identification o f simple and relevant indicators o f need andeffort. i.Increasingdataavailabilityandreliability. 3.91 Poverty targeting requires that information on poverty levels and on specific needs o f the target populations be available at the level o f each municipality. This means that municipalities should include, for instance, a social welfare department responsible for collecting and updating relevant data. This has been attempted in Chile with relative success.36 ii.Devisingasimpletransferformulatakingintoaccountbothneedsandeffort. 3.92 The objective o f general purpose grants is to ensure that each local government has the ability to provide a minimumpackage o f public services (beyond education and health) inaccordance to their constituencies' expressed preferences. Inorder to target the neediest, while at the same time maintaining an incentive for fiscal rigor, the transfer design needs to devise an efficient yet accurate way to measure the recipients' need for additional resources and effort inraising own revenues (based on capacity). 3.93 Inthe African context, need can often be captured only through very "rough" proxies such as demographic data weighted for specific situations. In Madagascar the weight factor could be, for instance, some measure o f "remoteness" or a classification o f communes based on socio-economic characteristics developed in the poverty map. In addition, the observed differentials inthe cost o f service delivery could be included. 3.94 Estimation o f fiscal capacity (not just actual revenue collection) and thus o f effort, requires a minimum o f available objective demographic and economic data. For transfers to communes, the formula should include a proxy that captures the communes' main source o f revenue. At minimum, some rough count o f taxable property and businesses would be needed. In the Malagasy context, existence o f or proximity to a market could also be considered as an important element. 36Onthe Chile example, see BirdIADB,TechnicalPaper Series, 2000. 45 3.95 To ensure maximum efficiency o f transfers and to maintain strong performance incentives, mechanisms of sanctions, and rewards can be imagined. The Uganda Local Development Grants provide a straightforward model. The basic idea is that grant allocations to each entity from year to year will be re-evaluated according a set o f performance indicators. While good performers see their envelopes increase by 20 percent, poor performers see a similar decrease in their allocations. To harmonize the recipients' capacity to perfom efficiently, the relevant ministry (in Madagascar the decentralization ministry) i s expected to play a strong coordination and mentoring role. Transparency can be fostered by allowing for regular inspections and timely publication o f all transfers allocations (planned and actual). 46 4. COMMUNES 4.1 Communes are Madagascar's local-level governments. They should not be confused with "communities", which are either (i) o f social and economic groups community organization (like fishermen, farmers', or parent-teacher associations) or (ii) villages (the Fokonolona). Since November 1995, after a first round o f municipal elections, communes constitute the lowest level and institutionally most advanced level o f sub-national government. 4.2 Madagascar has a total o f 1,392 communes, out o f which 45 are classified as urban and 1,247 as rural: 0 Rural communes are subdivided in (i) Primary rural communes (Communes rurales de premiBre catigorie), which typically represent a district headquarter, and (ii) Secondary rural communes (communes rurales de deuxisme catigorie), which correspond to the majority o f communes inthe country; 0 Urban communes are subdivided into three categories (i)the capital Antananarivo has a special status; (ii) the provincial capitals (including Antsirabt but excluding Antananarivo); and (iii) remaining 38 urban communes. the 4.3 The focus of this study is on rural communes, where 70 percent o f Madagascar's population and the large majority o f Madagascar's poor live. On average, rural communes have a little under 8,000 inhabitants; the smallest commune has about 2,000 and the largest rural commune has more than25,000 inhabitants. Many communes cover vast areas, sometimes more than 1,000 km2.37 4.4 This chapter builds on research and commissioned fieldwork o f the World Bank since 2001. After describing the institutional context, it will present the results o f in- depth research of local govemment finance, including needs for investments and recurrent expenditures at the commune level and the sources o f finance that currently meet parts o f these needs. The chapter will then present the revenue and expenditure composition in rural communes and compare this to budget data from selected urban communes.38 37The population data is based on the 1993 census and then adjusted annually with an average population growth. There seems to be a large underreporting of urban communes. A new census was planned for 2003, but may only be carried out in2004. 38A more detailed analysis o f inequalities and financial management performance among rural communes will be presented in a separate paper on local government finance in Madagascar. See Emanuela GalassoFrank-Borge Wietzke, Determinants of local government finance inMadagascar, forthcoming. 47 INSTITUTIONAL ARRANGEMENTS AT THE COMMUNELEVEL Communeinstitutionsandfunctions 4.5 Local administration has been exposed to several changes in its roles and responsibilities since the beginningo f the decentralization process in 1995. The first and most important turn came with the 1998 constitution, which gave the provincial legislature the mandate to redefine the status, competencies, and rights o f communes. However, with the inactivity o f legislative bodies at the province level after the 2002 crisis, communes remained de facto governed by largely outdated and often contradictory laws and decrees which were issued across the various phases o f decentralization policy inthe 1990~.~' Competencies of communes 4.6 Commune competencies are defined by a host o f different laws and decrees. In summary communes are responsible for (i)maintaining administrative services, (ii) managing the public domain and basic economic infrastructure, (iii) management waste andpublic hygiene, (iv) providing basic social services, (v) managing local security, and (vi) development planning. In addition to these general functions, communes are responsible for identifying and coordinating all local investments as well as supporting implementationo f the Government's PRSP at the local level: Administrative services. A core function of communes is to carry out a number o f administrative services. These include the (i) management o f the civil registry (Etat civil, including the issuing o f birth certificates, passport, registering marriages and name change, etc.), (ii)approval o f inhumations, customary exhumations, and transfers o f corpses (together with local health staff), (iii)formalization o f signatures and official documents, and (iv) assistance to central government agencies inmilitary drafting and the organization o f elections. Management of the public domain and economic infrastructure. Communes are the formal owners o f local public domains, such as communal roads, parks and places, sport grounds, assembly halls, swimmingpools, cemeteries, market places, slaughterhouses, latrines, town halls, and irrigation schemes. Specific responsibilities with respect to the public domain are poorly defined but generally include maintenance, security, and administrative functions. Only in the case o f economic infrastructure, public parks, and sport grounds are communes required to carry out small investments (formally 15 percent o f their budgets have to be allocated for this purpose). Wastemanagement andpublic hygiene. Like inmany other developing countries, communes are also responsible for waste management and sewage. In addition they have the right to establish aBureau Communal d'Hygi8ne whose functions are to monitor public hygiene (public space, houses, disinfections, rats, vaccinations, etc). Social services. Following the introduction o f conditional transfers for education and health, communes have become formally responsible for carrying out basic 39Laws 1994-01, 1994-06to 1994-08, 1995-005. 48 investments, maintenance, and procurements for public primary schools and health facilities. Expenditures are planned and executed in cooperation with local service staff. Recurrent salary expenditures such as for teachers or extended service staff (pharmacist, guards) are sometimes paid by the commune as counterpart payments for central government investments in health and educational infrastructure. In addition, communes are required to provide social safety net functions for vulnerable households at the local level. Security. Communes are responsible for municipal police functions (public order, regulation o f markets and festivities, traffic, migration control). These functions can be carried out by communal staff or, upon request, by regular police forces. The final decision on the use o f force remains with the Sous Prefet and the police command. Inmore isolated areas local governments can request the installation o f gendarmerie posts (Poste avarice?. However, this service has to be subsidized by the commune (accommodation and living expenses). In addition, communes can maintain a communal shelter WourriBre) inwhich they collect non-claimed animals and vehicles. This function often has a strong relevance in areas with high incidence o f cattle theft. Developmentplanning. Communes above 10,000 inhabitants are formally required to develop and implement urban development plans. These plans should identify future investments, zoning, plans for infrastructure and equipment, etc. Construction permits are provided by the commune on the basis o f its urban development plan. Local development planning was initially not mandatory for smaller rural communes but became a regular function with the introduction o f the Plan Communal de Development (PCD), which is inspired by the PRSP and outlines commune development priorities (see Box 4.1). Box 4.1: Local developmentplanning(PlanCommunaldeDCveloppement,PCD) Since 2001 all communes are formally required to produce communal development plans. These plans are supposed to be elaborated, similarly to the PRSP-process at the national level, in participatory processes involving local populations and civil society. The plans should outline local investment and development priorities. Donors have pledged to only support projects that are included in the communal development plans. Despite positive experiences inanumber ofcasesthe following drawbacks havebeenobserved: 0 Inmany casesthe PCDprocess hasbeenhighlydonor-driven andwas not alwaysbased on bottom-up decision making. 0 Particularly smaller communes do not have the capacity to engage in strategic planning or to provide good M&E o f their PCD. 0 In most cases, communes did not define and reserve funding for own investments, counterpart payments, or to cover maintenance costs. Institutional set-up 4.7 Madagascar's communes are formed by an executive bureau and a commune council. The executive bureau is headed by the mayor who i s directly elected by the commune population. There i s no formal organization structure for commune administrations. Yet, in general the town hall consists o f two deputies and a minimum o f three department heads (for finance, general administration, and the Secrktaire de 1'Etat 49 Civil).Many communes also include departments for cleaning, security, and engineering (technique). The commune council is elected through separate legislative elections. The council's main role i s to oversee and control the commune administration. The central administration is represented at the commune level by the DAA who directly reports to the Sous-Prefet, who inturnrepresents the key outpost of the Interior Ministry. Budget management 4.8 Budget programming and execution are hampered by structural over- programming o f revenues and expenditures. Formally each budget prevision should reflect revenues and expenditures o f the previous three fiscal years. Yet in order to bypass restrictions on budget reallocations, communes generally present highly overestimated revenue and expenditureprevisions intheir annual budget programs4' 4.9 The practice o f overestimating revenue and expenditure previsions is at odds with basic budgeting procedures. The general lack of funds and the late arrival o f transfers and revenue allocations further complicate strategic planning. As a consequence, most communes have ad hoc management styles, where budgets are executed according to the availability o f funds rather than on the basis of strategic priorities. Particularly large investments or seasonal increases in recurrent expenditures (i.e. related to repairs and maintenance o f communal infrastructure before or after the rainy season) are difficult to realize inthese incremental budgetingprocesses. Sub-communegovernments 4.10 Because o f the large surface area of most rural communes, most villages are quite distant from each other and also far away from the commune headquarters. In many cases, villages are more than 25 km away from the commune headquarters. These distances and the lack o f transportation make the sub-structures o f communes particularly important. On average there are 10 Fokontany (FKT) and 30 Fokonolona (FKL) per commune. The role of the FKT was upgraded and partially formalized at the end o f the 1970s. The FKL are sub-structures o f the FKTs, which represent the traditional unit o f Madagascar's social organization (dating back to the pre-colonial period o fthe Merinakingdom). 4.11 The performance o f FKT and FKL varies and depends on the local institutional context. Active sub-commune governments play a role in carrying-out some commune functions. They help to coordinate local affairs in villages that are far from communes, maintain public security (see Box 4.2), regulate formal and customary law, and supervise trade and transport in agricultural products. In addition, they support mayors in health and education campaigns, assist in administrative matters such as monitoring o f village population and the Etat Civil, prepare elections, and organize village meetings. FKTs 40 The budget process is divided into three principal steps, the presentation o f the budget primits (first previsions), the budget additionel (revised previsions), and the compte administrative, which presents the executed budget. Reallocations o f expenditures are possible between the budget primitif and the budget additionel. However, because this option requires lengthy approval procedure by the commune council and deconcentrated control agents it i s seldom practicedby many communes. 50 and FKLs also manage allowances for administrative representatives and security forces, as well as collect local taxes and user fees (Vonodina). 4.12 Despite their important functions for communal service delivery, the roles and institutional relationships o f FKT and FKL remain an area o f contention. The tension around the status o f FKT and FKL, and their relationship to the commune is heavily influenced by the broader disconnect that exists between deconcentrated control agents (Sous Prefets and DAAs) and decentralized local governments. The lack o f clarity relates in particular to the (i)horizontal relationship between the commune administration (headed by the mayor) and the deconcentrated administration (represented by the DAA); and the (ii) vertical relationship between the commune and sub-commune structures. 0 Horizontal relationships: Since the local elections o f 1995, all communal affairs should have been managed by the decentralized administration, headed by the mayor. Yet, in practice, the DAA and other deconcentrated agents continue to carry out executive functions like the collection o fmost local revenues; 0 Vertical relationship: Since their formal establishment by the Second Republic (1975-1992), villages are part o f the deconcentrated command chain (with the head of the village appointed by the DAA). Since the constitutional changes o f the third Republic, villages became sub-units o f the communes. However, the village heads continue to be nominated by and accountable to the DAA and the Under-prefect. 4.13 The lack o f coordination between deconcentrated and decentralized reform strategies is also reflected ina November 2002 govemment decreethat seeks to formalize the informal governance structures o f the FKTs and FKL (decree 2002- 1170). Under this decree, both FKTs and FKL report directly to the Sousprefets and DAAs, acting as their local representative. Sousprefets have the right to nominate and fire FKT presidents as well as to establish administrative demarcations o f FKTs and FKL. FKL presidents are elected through community assemblies. However, Sousprefets and D A A s can decide on the suspension o f FKL executive organs. 51 4.14 The decree can be seen as a valuable attempt to fill the existing regulatory vacuums at the commune and sub commune level. Moreover it seeks to formalize the use o f customary law at the village level (Dina) and strengthens the competencies and discretion o f FKTs and FKLs in carrying-out local police functions - an attempt o f central authorities to address the declining security situation many regions o f rural Madagascar. The decree goes as far as to entrust local security groups within the FKL with the discretion to use arms (excluding fire arms), to monitor cattle theft and to pursue alleged criminals. 4.15 However, with its strong focus on deconcentrated command lines this innovation risks to jeopardize recent gains in decentralization to communes. Communes are threatened to become an empty shell o f local govemment because they now have to establish collaboration with FKT / FKL through the deconcentrated command chains, in particular the Sous Prefets (see chart 4.1). This i s problematic because the decree also assigns FKT and FKL broad functions in areas that traditionally fall under the competencies o f communes such as municipal planning, economic and social services, publichygiene, and security. Chart 4.1:Institutional relationships at the local level DECONCENTRATED DECENTRALIZED DISTRICT '.. ..-.. ________--- ..3 Mayor COMMUNE Etatcivil ;Local security Maintenance o f social and economic infrastructure Local development planning MEASURING COMMUNE FINANCE: SAMPLE AND METHODOLOGY 4.16 The backbone o f the evaluation is two sets o f data and analysis: (i) in-depth an analysis in 15 rural communes evaluating all flows o f funds and estimating local financing gaps, and (ii)a data set o f 321 rural commune budgets allowing for a quantitative evaluation o f uses, sources and determinants o f communes' own finance. (i) The 15 communes study estimated the localfinancing gap based on two inputs: (i)assessmentofunmetbasicserviceneedsinthehealth, education, andwater sectors, and (ii)combination o f four sources o f service finance: a) commune 52 budget (which includes central government transfers), b) deconcentrated financing, c) community contributions, and d) donors (see Annex 5 for a detailed description o f the methodology). Despite the small size o f the sample, the results for commune budget data proved to be robust compared to the data set o f 321 communes. Financing through central ministries and community groups i s likely to vary across communes as much as communes' own resources. The largest deviations can be expected from donors, who provide the largest and least systematic source o f financing. (ii)provided The 321 communes data set, which was collected at the ministry o f budget, the basis for a quantitative analysis o f communes' revenues and expenditures. Given the large number of non-reported budgets the sample had to be tested for representativeness (see Annex 6). This confirmed that the sample was largely representative for rural communes, even though there was an underreporting o f the most isolated and smallest communes. Chart 4.2: Data Sources 15 commune study Commune Deconcentrated Community Donors Budget Spending contributions 321 commune data set 4.17 In addition, aggregated budget data was retrieved from six urban communes across the country as well as three sub-urban communes in the area around the capital Antananarivo. This data was used to establish estimates on revenue and expenditure trends in urbancommunes. Direct comparisons proved to be difficult to establish due to different budgetnomenclatures inthe two data sets. COMMUNE NEEDSAND OVERALL FLOWS OF FUNDS 4.18 With an annual per-capita income of around US$240, Madagascar is one o f the poorest countries in the world. Given the strong divide between rural and urban poverty rates, needs can be expected to be very high in rural communes. The magnitude o f under-financing o f basic local services, as measured by the financing gap analysis, is consistent with Madagascar's poverty rate inrural areas. 4.19 However, the challenge o f filling public services needs often goes beyond the challenge o f providing financing. As illustrated in Box 4.3, for the education sector in the districts o f Andilamena and Farafangana, the degradation o f public schools, and widespread insecurity have caused the closing o f schools and the expansion o f 53 community-managed schools. The lack o f monitoring, evaluation, and beneficiary participation creates dissatisfaction about CISCO management. Box 4.3 :Case studv of the schoolsinthe schooldistrictof FarafanganaandAndilamena Source: Miara Mita 2003 Needs: The local financing gap 4.20 Existing service finance generally does not suffice to satisfy unmet service needs. This is particularly true in the education sector where a financing gap o f US$16.2 per capita indicates substantial unmet needs. The gap inthe water sector appears smaller, but this reflects the large amounts o f donor investments in a small number o f sampled communes. However, absolute financing needs in the water sector remain substantial with US$8.8 per capita (see table 4.1). 4.21 As table 4.1 shows, service performance i s uneven across key social sectors. Access rates for infrastructure are particularly low in education. Even the conservative estimate, based on minimum infrastructure requirements and covering only those children already in school, points to significant investment needs (US$15 .O per capita). Water needs are also considerable (US$13.8 per capita). In the health sector, where the service network is better developed, investment needs are comparatively lower at US$4.6 per capita. Financing needs for recurrent expenditure in both social sectors are comparatively manageable and more evenly distributed(below US$2.5 per capita). Table 4.1: Localfinancinggaps (US$ per capita and YOoftotal needs) 54 4.22 The combination o f all flows still results in very modest per-capita spending. In the health and education sectors, (i) per capita investments were less than US$1, average and (ii) per capita recurrent expenditures amounted to US$1.60 (education) and average US$0.65 (health), most of which for salaries (which cannot be reallocated to meet additional needs). Non-salary recurrent expenditures in the two social sectors are even lower - around US$0.35 per capita. Financing for water and sanitation is significantly higherwith almost US$5 per capita because o f above average donor investments inthree ofthe sampled communes (see Chart 4.3). Chart 4.3: Gap per sector (US%per capita) Education ' Health 0Water Source: Miara Mita, Commune Finance Study 2002 4.23 Ifthe financing gap results were tentatively extrapolated to the whole country, Madagascar's rural communes would need resource transfers in the magnitude o f US$320 million to cover unmet recurrent and investment needs in the three sectors. In 2001, US$320 million represented 7 percent o f GDP, 43 percent o f Madagascar's budgetary resources and forty times the current level o f annual block transfers to rural communes. This presents a conservative estimate as (i) it only covers minimum sector norms and does not include important other priorities such as security or transport, and (ii)coverage hadbeen limitedinthe education sector. Compositionof availablefinancing 4.24 The relative importance o f different financing sources varies significantly between capital and recurrent expenditures, as well as between deconcentrated and local sources of finance (see Table 4.2): 0 Donor investments represent the single most important source o f finance for capital expenditure inall three sectors. 0 Line ministries (health and education) represent the second biggest source o f local service finance with respectivelyUS$0.40and US$1.3 1per capita, which i s mostly spent on salaries. 0 Districts only play a minor role with the exception o f investments in the education sector (US$0.13 per capita). The value o f total spending i s much 55 bellow the 406 million investment subsidy, which would represent an average per capita investment o f more than US$0.50. 4.25 Community groups' contributions were found to play a stronger-than-expected role for service finance in all three sectors. User fees and cost recovery schemes in the health sector sum up to US$0.21 per capita, accounting for almost one third o f total recurrent expenditures. Likewise, communities also take in charge salary expenditures such as in the education sector, where teachers are often directly employed by local parent-teacher associations. 4.26 Local governments provide only a marginal share to overall service finance in the three key sectors. Almost 70 percent of commune expenditures are spent on local administration (see chart 4.8). Nevertheless, these expenditures translate into some important administrative and organizational services not captured in the sector gap analysis. These administrative functions affect public service delivery and financing. For example, birth certificates delivery i s crucial as school children must have formal birth certificates to graduate from primary school (see below). Table 4.2: Sources of Financing (US$ per capita) Source: Miara Mita, Commune Finance Study 2002 COMMUNEREVENUES Local revenue assignment and collection 4.27 In Madagascar, revenue assignments roughly comply with standard decentralization theory. They have been further simplified since the introduction o f the 100 percent allocation principle for each tax, assigning revenues to one level o f government only (see Chapter 3). The most important revenue sources o f communes include: 0 Property taxes on land (ET) and buildings (IFPB), as well as an additional tax on buildings, which serves as a conditional tax to finance commune counterparts to donor investments (TAFB); 0 Excise taxes and licensing rights on gambling machines, festivals, alcoholic beverages; 56 0 Taxes on localnaturalresourcesand agriculturalproducts ("ristoumes"); 0 Administrative and user fees like charges for the utilization of communal infrastructure and public territory (see Annex 8 for a full compilation o f all commune revenues). 4.28 Despite the assignment o f these taxes to the local level, true local revenue mobilization i s hampered by (i) the central determination o f revenuerates, and (ii) collection o f key revenues by deconcentrated intermediaries: (0 Determination of revenue rates: In most cases, tax levels and user fees are determined centrally, notably for the property tax, mining taxes, and a number o f user fees, leaving local govemments only with only a small margin (in the case o f the property tax 2-5 percent) indeterminingtheir own desired income levels. (io Revenue collection: While revenues are appropriately assigned, collection is distributed across several decentralized and deconcentrated institutions: the commune, the DAA, the Treasury, the DGI, and the FKT. Most important local taxes such as the property tax and miningtaxes, as well as a number of administrative fees, are collected by central and deconcentrated agents such as the representative o f the treasury and the DAA (see Table 4.3 andAnnex 8). Revenuesource/item Collectionmechanism Property tax Treasury (exception: 12pilot communes under PAIGEP) Excise taxes DAAor "administration fiscale" Local natural resources ("ristournes") FKT, commune andDAA I Administrative and user fees I Mostlv commune I The rural-urbandivide 4.29 Commune finance in Madagascar is characterized by a great divide between urban and rural revenue mobilization performance. While urban communes have increased their own revenues several-fold since the mid-1990s, rural communes' revenues remained marginal both in absolute and in per-capita terms. As a result, on average 72 percent o f urban communes' revenues are own revenues and transfers are only28 percent. The only exception i s the commune o f Fianarantsoa, where transfers represented more than 50 percent o f total revenues. In rural communes, the situation is the opposite: own revenues finance one fourth o f the budget, while three quarters come from transfers. Semi-urban communes in greater Antananarivo stand inbetween with 47 percent of own revenues and 53 percent transfers (see Table 4.4). 4.30 The rural-urban divide is also visible in the composition of revenues. Urban communes have now a stable base o f tax revenues and user fees, while rural communes -inparticularlythepoorest-dependto alargedegree onfluctuatingrevenuesources, such as sales taxes on local primary goods. The rural-urban divide is particularly visible in property tax collection, a local tax with potential as demonstrated by Madagascar's urban communes. Starting from a low base, these revenues can grow fast. However, in 57 rural communes, economic, administrative and logistical constraints have inhibited similar performance: 0 Urban communes such as Antananarivo, Tamatave, Majunga, and Ambalavao registered significant increases in the property tax following the intervention o f the French Cooperation's technical assistance project PAGU. According to P A G U data, the transfer o f responsibilities for property tax collection from DAAs to communes along with updating property cadastres improved property tax recovery rates in 10 urban communes by more than 50 percent in the first year of their i n t r o d ~ c t i o n . ~ ~ 0 Rural communes, by contrast, saw very little improvement inproperty tax yields with the exception o f some improvements in 12 less isolated communes targeted by the World Bank's project PAIGEP. The main constraints include (i) insecure land rights and outdated valuations, (ii) unclear responsibility on collection (see below); and (iii)higher institutional level constraints, like inadequate enforcement mechanisms (police administrative courts) or delays o f approval o f tax emissions. Table 4.4: Revenuecompositionof urbanand rural communes (estimated) Urbancommunes Rural communes Greater Antananarivo National average -I_ o f own revenues 72% 47% 25% o f which: Tax revenue 34% 17% 12% User fees 28% 27% 11% I Share Others 1 10% I 3% I 2% I o f transfers 28% 53% 75% I I I Note: Urban communes included in the sample are Antananarivo, Fianarantsoa, Majunga, Tulear, Tamatave, Fenarive Est, Ambalavao. Averages are based on budget information for the years 1999-2000, except for Majunga (1999-2000) and Fenarive Est (2000-2001). Communes in the greater Antananarivo area include Ambohidrapeto, Tanjombato Andranonahoatra (year 1999 only). The national average for rural communes calculated based on the data from the sample of 321 communes. Shares of revenue composition are calculated on the basis of recettes de fonctionement. 4.31 A last dimension o f the rural-urban divide is in the transfer system. Although they have a stronger own-revenue base, urban communes paradoxically receive almost twice the amount in per capita transfers than rural communes (US$1.54 compared to US$0.86 inrural communes in2000).42 As a result, urban communes receive 25 percent o f total transfers with only 15 percent o f the population. Rural communes are penalized even though their degree o f poverty i s higher and their revenue raising capacity lower (see chapter 3 on transfers). 41 Source: CooptrationFranGaise. 42 However, real per capita amounts may be lower if we assume that underreporting inMinistryo f Budget populationdata is stronger inthe case o f urban than rural communes. 58 Box 4.4. Social funds and decentralization -movingprogressively away from the rural-urban divide? In a country like Madagascar where there are strong differences in the administrative and financial capacity o f rural and urban communes, social funds can offer an important alternative to broad based devolution o f investment and service delivery functions. Yet, in the past social funds also weakened local capacities, as they tended to bypass local authorities and deconcentrated sector administrations inthe planning and implementation of their investments. To address these shortcomings, social funds across the world have begun to shift their attention from simple investment finance to broader objectives o f local institutional reform and capacity building.Madagascar's social fund project FID (Fonds d'Interventionpour le Developpement) now coordinates all local investments with local authorities and trains commune govemments to take an active role in the coordination of local planning processes (PCD). In addition FID has begun to channel investment hnds directly to local governments, involving them closely in financial management, procurement, and supervision o f construction works. The fund also provides operational support to ensure better service delivery once an investment i s completed (i.e. training o f commune staff, user groups and school teachers). The social fund's approach to engage communal administrations in the management and operation o f local infrastructure projects offers a good altemative to fell full fledged devolution of public services as it allows local governments to gradually build up their administrative and financial capacities. Government and donors alike are therefore exploring ways to improve FID's outreach to local govemments, including the scaling up and deepening o f training programs to local governments and improving the integration of FID project funds into commune budgets.43 I Sources:Barrientos 1999;Parker & Serrano 2000 Rural revenue performance and composition 4.32 Rural communes' o w n financial resources are extremely small (equivalent o f US$0.20 per annum) and will continue to play a marginal role. Even when conservative population estimates o f the Ministry o f Budget were applied, average per capita revenue of rural communes only amounts to US$1.12, which in turn is dominated by transfers (US$0.86 per capita). A s a proportion to aggregate financing, communes own finance represents less than: 0 2 percent o f total government revenue; 0 10 percent of total local finance (which also includes deconcentrated, communities and donors; see also Annex 5); 0 5 percent o f total local finance inrural communes.44 4.33 Rural communes' own revenues are composed o f three principle sources: (i) tax revenues, (ii)administrative fees, and (iii)user fees. About 7 percent o f commune 43 At this time FID funds are still channeledthrough specialbank accounts inthe name o fthe communes. 44 Source: Miara Mita. 2003, table 3.2). 59 income are covered through various other sources such as savings, budget overhangs from the previous fiscal exercises, alienation o f assets, grants anddebts (see chart 4.4).45 Chart 4.4: Breakdownof total local revenues Others Transfers 74.8% Source: Bank staff calculations based on Miara Mita, Commune Finance Study 2002 4.34 Rural communes revenues are biased towards productive activities (mainly agriculture and livestock) and commerce. Within tax revenues, half o f the resources come from agriculture related economic activities. Inaddition, the sale, registration, and slaughtering o f livestock constitutes a substantial component o f administrative fees. Most o f the revenue from user fees also related to markets and communal infrastructure (see Charts 3.5, 3.6, and 3.7). 4.35 In contrast, classic sources of local revenue such as property or business taxes play a rather marginal role in the majority o f rural communes. A large share o f the sampled communes report income form the property tax, but this tax source still accounts for a small fraction (13 percent of taxes) o f the communes' tax revenue. The same goes for business taxes, which only represent 4 percent o f local taxes. Rural communes have very few formal businesses and business taxes are thus collected less frequently than inthe sample average. Consumption and excise taxes (such as on sales of alcohol, local festivities, or gambling) are comparatively more important and represent an average share o f 13 percent o f owntaxes inthe sample total (see Chart 4.5). 4.36 Like in many other poor developing countries, Madagascar's rural communes systematically under-exploit tax revenues, in particular the property tax. As a consequence, most rural communes rely on alternative sources o f local revenue. Larger communes receive important shares o f revenue through user fees for communal infrastructure (markets, housing, tourist sites, etc.). Smaller rural communes that do not dispose o f comparable infrastructure endowments have to rely on economically more counterproductive and volatile revenue sources such as sales taxes on local primary goods. 45 Commune debts also form part of this category, typically represented by advance payments from the treasury or arrears o n water and electricity bills. However, given that communes are not allowed to run budget deficits,this residual category is reportedirregularly andplays only a minor role. 60 Chart 4.5: Breakdownof total localtaxes Chart 4.6: Breakdown of user fees I business fisherylforest agriculture 0% 5% 10% 15% 20% 25% 30% $59 I Source: Bank staff calculations based on: Miara Mita, Commune Finance Study 2002 Chart 4.7: Breakdown of administrative fees admilegal 0% 10% M% I%YIX 40% MX 70% I Source: Bank staff calculations based on: Miara h ita, Commune Finance Study 2002 4.37 One o f the most important findings that came out o f the analysis o f local government finance concerns revenue- collection mechanisms. Madagascar's deconcentrated intermediaries are much less effective in revenue collection than local governments themselves. Deconcentratedand agents o f the central state, who collect the majority o f local taxes and fees, account for less than 30 percent o f local revenues inthe sampled In contrast, the few revenues that are directly perceived by the commune account for almost 70 percent o f local governments' own income (see Table 4.5). User and administrative fees constitute the most important share within the revenue that i s collected directly by the commune (about 35 percent). Even more telling, communes collect significantly higher shares o f local taxes (26 percent) than their central and deconcentrated counterparts (20 percent). This finding illustrates that, despite all capacity constraints, communes display a certain readiness to take over more complex revenue collection responsibilities than i s currently the case. 46The exact shares are difficult to identify because institutional arrangements for revenue collection often differ at the local level. 61 Table 4.5: Own revenue by recovery mechanism % reportinglshare of total own Std dev I________--_____.-- item revenue ____.___I___--- Communes 96.9% 69.3% 0.24 administrativefees 93.5% 14.5% 0.16 user fees 84.4% 20.1% 0.21 Taxes 93.8% 26.4% 0.25 Ristournes 53.3% 12.4% 0.19 Deconcentrated agents 96.9% 27.9% 0.21 administrativefees 79.4% 7.9% 0.13 Userfees Taxes 96.3% 20.0% 0.19 Ristournes 34.6% 6.7% 0.14 Others 9.4% 0.4% 0.02 COMMUNEEXPENDITURES Urban versus ruralexpenditure 4.38 Urban and rural spending patterns are more similar than their revenue composition would suggest: Recurrent expenditures represent a high share o f spending in both urban and rural communes (around 90 percent). However, the largest urban communes tend to finance investments through external loans and donor projects, which artificially increases the share o frecurrent expenditures; Difference remains in the share o f wages, which still consumes a disproportionate share of rural commune spending (60 percent) compared to 37 percent in urban communes; Estimates for functional expenditures, indicate that spending on administration i s similar (at around 70 percent) but urban communes spend significantly higher shares o f their budgets on economic services (32.4 percent), while spending on social services remains much below the trends for rural communes. This trend i s possibly explained by the fact that urban communes spend important amounts on economic services such as street cleaning and waste management, while social expenditures - notably in education and health - are made directly to central and private service providers. 4.39 An interesting trend can be observed inthe case o f the three rural communes in the suburban belt around Antananarivo. These communes sometimes spend significant amounts on investments (the average is 37.8 percent, but the range goes up to almost 70 percent inthe case o f Tanjambato). Most o f this investment goes into administrative and economic infrastructure, while social investments only account for a marginal share. Economic expenditures are less important than in larger urban communes, but still 62 significant with an average share o f around 23 percent. Social expenditures account in total for 8.4 percent o ftotal expenditures (see Table 4.6). Table 4.6: Breakdownof expendituresinurbancommunes and communesinthe greater Antananarivo area Urban Greater Rural communeshtananarivo communes share investments 7.4% 37.8% 10.0% wages as share o ftotal expenditure 36.8% 13.9% 66.1% wages as share o frecurrent 38.1% 20.3% 59,3% social expenditures / share o f total 1.9% 8.4% 15.5% economic expenditures / share o f total 32.4% 23.0% 14,5% administrative expenditures in% o ftotal 65.7% 68.6% 69,8% Note: Urban communes included inthe sample are Antananarivo, Fianarantsoa, Majunga, Tulear, Tamatave, Fenarive Est, and Ambalavao. Averages are based on budget information for the years 1999- 2000, except for Majunga (1999-2000). Communes inthe greater Antananarivo area include Ambohidrapeto, Tanjombato Andranonahoatra (year 1999 only). The national average for rural communes calculatedbased on the data from the sample o f 321 communes. Ruralcommunes' spending 4.40 Looking at the finctional and the economic breakdown o f commune expenditures, the following trends stand out (see Chart 4.8): 0 Communes' spending i s dominated by recurrent expenditures (90.0 percent) while investments only represent a marginal share in commune budgets (10 percent o f total). Recurrent expenditures are mostly spent on wages and related expenditures (almost 60 percent o f total expenditures). 0 Dominance o f administrative expenditures (70 percent) over economic (14.5 percent) and social (15.5 percent) spending. Chart 4.8: Economicandfunctional expendituresof ruralcommunes 1 80% 1 74.0% 75.5% 70% 60% -2 50% E e 42.2% Investment 40% EWage y030% 0non-wage 20% 10% i 1 7.1 14,5%% 0% t 1 Administrative Economic Social Functional Source: Bank staff calculations based on: Miara Mita, Commune Finance Study 2002 63 Recurrent expenditures 4.4 1 Notwithstanding the overall importance o f wage-related expenditures by communes inMadagascar, most of the wage expenditure is largely accounted for by the allowances to executive and administrative staff and to commune council members (3 1.8 percent o f total expenditures). A smaller fraction of the total budget is used for permanent and nonpermanent wages, which are more directly related to the provision of public services inthe communes.47 Table 4.7: Breakdown of recurrent expenditures % communes Mean budget share reporting item (whole sample) Std dev Total wage 99.1% 59.3% 0.16 Permanentwages 97.8% 22.3% 0.11 Non permanentwages 77.6% 5.1% 0.06 Allowances 99.1% 31.8% 0.13 Total non-wage 99.1% 30.4% 0.13 Debt service 21.8% 1.9% 0.05 Equipment 96.0% 4.4% 0.03 Subventions to others 81.9% 8.5% 0.08 Maintenance 79.75% 5.23% 0.07 Transport 71.34% 1.66% 0.02 Others 95.95% 8.77% 0.06 Source: Bank staff calculations based on: Miara Mita, Commune Finance Study 2002 4.42 The beneficiaries of allowances are difficult to identify on the basis o f the budget nomenclature alone (there are many non-specified allowances. This explains the high share o f the category "others" in table 4.7). However, with the limited information available, it is possible to attribute about 5 percent o f total allowances to members o f the executive o f the commune (mayors and deputies) and 26.7 percent to council members. This illustrates that - while generally important - horizontal control through the commune council can impose important financial burdenson the budget o f the commune (particularly inthe case o f smaller communes). Investments 4.43 Capital expenditure o f communes accounts on average for 10 percent. This is less than the legally required 15 percent o f total expenditures that should be allocated to investments. However, there i s quite some variability across communes because only 67.9 percent in the sample report investments. Therefore capital expenditures in the communes that do invest represent a much higher share, reaching almost the required 15 percent threshold (14.7 percent on average). Within capital expenditures the biggest share i s allocatedto construction andrehabilitation works (see table 4.8). 47Non-pennanent staff i s hired for additional requirements such as cleaning drainage systems before the rainy season etc. Other services related to public hygiene, notably the fight against parasites and other natural plagues are generally taken into charge by the communes. 64 Table 4.8: Breakdownof investments average budget % communes share (whole reporting sample) Std dev total investments 67.9% 10.0% 0.14 Investment (if reported) 15.7% 0.15 formal investmentallocation (15%) 10.0% 1.3% 0.07 Breakdown by type: Purchase of assets 20.6% 1.O% 0.03 Construction/rehabilitation 63.2% 9.0% 0.13 Source: Bank staff calculations based on: Miara Mita, Commune Finance Study 2002 Spending- for administrative, economic and social services 4.44 The functional classification o f commune expenditures (administrative, economic and social) permits to assess the degree to which communes carryout their legally defined functions in the areas o f administrative, economic, and social services. However, the current budget nomenclature for local governments does not contain a clear functional classification. As a consequence the following analysis is based on a reclassification o f commune expenditures which was developed for the purpose o f this analysis.48 4.45 Commune expenditures are dominated by administrative services (69.7 percent o f total) while social and economic expenditures each account for roughly 15 percent o f total expenditures (see Chart 4.8 and Table 4.9): Administrative expenditures are dominated by the operating costs o f the town u ( 2 2 percent) as well as a large portion o f diverse administrative overheads that are difficult to attribute to specific functions (35 percent); Economicexpendituresare more irregular. The most regularly registered itemi s security, which i s reported by more than two thirds o f the communes. Security spending includes subsidies to accommodation and living expenses o f local police forces or salaries to local security forces (quartiers mobiles). However, in aggregate terms, spendingon security i s low with an average share of 2.7 percent of total expenditures in the sample total and 4 percent in the communes which report this item. Expenditures on waste management and public hygiene are reported by a good half o f the communes, yet overall spending i s even lower with only 0.8 percent o fthe total budget. Social expenditures account for 15.5 percent o f the total. Within social expenditures spending on social assistance - one o f the formal responsibility o f local governments - is insignificant with only 14 percent o f communes reporting this item. In contrast spending on education and health i s reported much more regularly (80 percent and almost 83 percent o f the communes respectively). 48 Classification errors may have occurred, particularly in the case o f cross-cutting expenditures (i.e. technical staff, equipment, and transport that may be used to support administrative, social, or economic services o f the commune). Incases o f doubt, we attributed all administrative overheads to ahnistrative expenditures andtechnical expenditures to economic services. Therefore there may be an over-reporting in administrative and economic services. 65 Spending on these services is very low with around 5 percent o f total expenditures spent on each o fthe two services. Table 4.9: Functional classification- overview % communes budget share reporting (whole sample) Std dev Administrative total 99.1% 69.8% 0.16 General administrative 99.1% 35.6% 0.15 Commune administration 99.7% 22.2% 0.11 Commune council 98.8% 8.1% 0.04 Executiveoffice 86.3% 2.5% 0.04 Revenue collection 77.0% 1.4% 0.02 Economic total 94.1% 14.5% 0.13 Others 72.0% 6.0% 0.11 Security 69.2% 2.7% 0.04 Transport 40.5% 2.2% 0.07 Markets 32.4% 1.5% 0.04 Water 29.0% 1.O% 0.03 Waste managementlHygiene 49.2% 0.8% 0.02 AgricuIture 13.7% 0.4% 0.02 Social total 96.3% 15.5% 0.10 Health 82.9% 5.5% 0.05 Education 80.1% 5.0% 0.05 Others 91.O% 4.6% 0.05 Social assistance 14.0% 0.4% 0.03 Source: Bank staff calculations based on: Miara Mita, Commune Finance Study 2002 4.46 If we compare social expenditures to conditional transfers for education and health we find that parts o f these services are financed through the commune's own funds. In 65 percent o f the sampled communes we observe that conditional transfer payments are lower than non-salary social expenditures. If we include salary related expenditures in the social sectors this ratio increases to 85 percent o f the communes. This illustrates that the majority o f rural communes utilize conditional transfers reasonably well. IMPROVINGSERVICEDELIVERYAT THE COMMUNE LEVEL 4.47 Communes should be the core o f the government's decentralization strategy, but there should be no illusions as to what communes can reasonably achieve. So far, decentralization to the commune level has not brought about the desired effects on the financial capacity o f local governments to support local service delivery. Although communes have been in place for almost seven years, they remain a marginal player in terms o f finance and local service finance and delivery. Commune finance represents less than four percent o f total government finance and even less in rural communes, where the large majority o f Madagascar's poor lives. Local services are mainly carried- out by the deconcentrated administration (for recurrent expenditures) and donors (for 66 investments). Rural commune financing i s even inferior to contributions o f informal community groups, which provide a growing share o f local service finance. 4.48 In order to improve the service delivery performance of communes, local governments should be empowered to carryout communes' functions, while central government agents should play a supervisory, supporting and controlling role. Co- administration at the local level proved to be inefficient, created tensions and contributed to a lower level o f revenue mobilization. It is thus important to implement (i) institutional reforms to streamline the institutional relationships; and (ii) reforms fiscal to put the communes hlly in charge o f local revenue collection while improving the execution o f transfers. (i) Institutiona1reforms a. Institutionalrelationships:The communeshould be in charge 4.49 Strengthening sub-commune structures (Fokontany and Fokonolona) i s laudable and important. Fokontany and Fokonolona committees (as well as other local institutions) can play important roles, both in strengthening downward accountability o f local governments and in carrvinn-out commune functions. In particular, strengthening sub-commune structures could help to better provide administrative services in remote villages. For instance, insufficient procedures to obtain basic registry services (like birth, death, and marriage certificates) causes major social problems for poor families as ex- post registration i s estimated at a prohibitive US$45 (FMG 300.000). 4.50 However, the detachment o f sub-commune structures from the communes (as it i s currently envisaged) carries the risk o f undermining effective provision o f communal services. Instead, it would be critical to establish clear reporting relationships between sub-commune institutions and communes and to limit the role o f central government intermediaries (Sous-Prefet and DAA) to control, supervision functions. Forthcoming analytical work on local horizontal accountability networks and institutional capacity building needs at the sub-commune level will help to formulate Bank support in this direction. 4.5 1 The deconcentrated agents o f the central administration, in particular the DAAs, should focus on their control and supervisory roles and where necessary provide technical assistance to local governments. This would simplify the roles o f deconcentrated agents at the commune level, strengthen the financial and legal control framework, and relieve the local governments from co-administration and central government intervention. b. Classificationand organizationof communes 4.52 The classifications o f communes in rural, urban, and firsthecond category are outdated because they do not capture differences in commune functions and recent changes inthe population size. The following changes are suggested: 0 The second category o f rural communes should be divided into sub-categories in order to account for differences in population size, economic potential, and degree o f isolation. 67 0 The classification o f second category urban communes and first category rural communes should be reviewed in order to make sure that the two categories correctly reflect differences inpopulation size and degree o f urbanization 4.53 In addition, the potential of inter-communal associations should be revitalized. Lessons from the well functioning inter-communal associations along the railroad Finanarantsu-Manakara and the Antananarivo suburbs demonstrate the utility o f such functional associations. Given the difficulty to systematically set-up inter-communal association, donors should continue to support them inthe medium-term. (io Fiscal reforms 4.54 For decentralization to work, communes need to increase their revenues. Otherwise the financing gap will continue to be substantial and incentives for efficient local service delivery hampered. There i s no magic recipe which would increase commune revenues to a level which would substantially close the gap. However, given the low level o f revenues in absolute terms, rural communes resources - own revenues and transfers - could be substantially increased inrelative terms. The following reforms would contribute to such an increase. a. Revenue collection 4.55 Communes should be in charge o f raising all o f their own revenues. Until now, three deconcentrated (DAA, Treasury, DGI) and two local institutions (Communes and Fokonatany) are involved in collecting local taxes. As demonstrated and as expected from accountability principles, communes are more efficient in collecting their own revenues than deconcentrated intermediaries. Eventhough communes collect only a few taxes and some administrative fees directly, these revenues account for almost 70 percent o f local governments' own income. Deconcentrated agents are still in charge o f collecting the majority o f local taxes and fees, but they only collected 30 percent o f local revenues inthe communes sampled inthis study. b. Revenue mobilization: the propertytax 4.56 To increase the potential o f property tax collection, which could represent one o f the most important revenue sources o f communes, technical assistance needs to be combined with participatory local census methods. Experience from a Bank-funded technical assistance project, PAIGEP, indicates that the constraints o f property tax collection are typically not overcome through sophisticated evaluation methods such as satellite or aerial photographs. Instead, participatory assessment procedures can significantly improve yields and reduce costs. Most promising results can be achieved whenpropertyvaluation is linkedto aprocess o f formalization o fproperty ownership. 4.57 In addition, a higher level of institutional support needs to accompany the property tax reforms. In the case o f PAIGEP, formal approval o f tax emissions was delayed due to insufficient involvement o f central and deconcentrated revenue administrations in the process. In addition, clarification o f legal regress procedures and strengtheningof provincial or regional courts would have been required to facilitate the enforcement o fproperty tax emissions. 68 c. Transfers 4.58 The foremost priority is to ensure predictable disbursements within a scheduled framework. Rural communes suffer most from the existing delays because transfers compose 75 percent o f their budgets on average. There needs to be a system for ensuring that the predefined amounts arrive in a timely fashion. For instance, the transfers could be split in two equal tranches to be disbursed every March and September, respectively. Predictable disbursements would allow communes to actually plan their expenditures and stop the existing practice o f ad-hoc budgeting. In addition, existing bottlenecks between the Ministry o f Decentralization and the Public Treasury would have to be removed. 4.59 The transfers to rural communes should be increased to help guarantee that rural communes have the minimumresources to function better. Rural communes should get at least the same per capita allocation as urban communes (FMG S,SOO), which would also remove the artificial and regressive separation between them. The increase o f transfers to rural communes would be fiscally neutral if the transfers to the districts (Fivondruna) were permanently suppressed. Transfers to districts have not translated in the equivalent local investments for which they were foreseen. 4.60 Communes and communities should be encouraped to gradually take over more far reaching competencies for service delivery in the medium term. While this paper recommends that local service delivery in the key social and economic sectors should remain in the hands o f deconcentrated administrations for the time being, there i s a strong potential to empower communes and communities to gradually take over the financing, planning, and supervision o f local services. At this stage, communes are already starting to coordinate public investmentsat the local level (Plans communales de developpement). Ina addition, a number o f projects, such as the social fund, have begun to execute local investments through commune governments. As these programs evolve, government and agencies should consider more systemic support to commune-executed investments. 69 c ANNEX 3 REVENUEASSIGNMENTINTHE2000 AND 2001BUDGET Loi de Finance2000 Loi de Finance2001 Excisetax Excise tax Importtax on petroleumproducts (TUPP) Import tax on petroleum products(TUPP) Stamp tax Stamp tax Tax on motor vehicles (vignette) 74 ANNEX 4 DECONCENTRATIONOF EXPENDITURES Table 1.Level of Deconcentration by Ministry, Recurrent Expenditures Center ProvincesRegions DistrictsOthers MINESEB 31.14% 55.93% 0.03% 12.90% I__ SANTE 45.83% 39.66% 0.34% 14.16% PECHES 73.34% 19.63% 7.03% AFF.ETRANG. 78.63% 21.37% (ambassades,consulats) JUSTICE 83.90% 12.62% 2.50% 0.97% (PresidentTribunal 1,2) TOURISM 83.95% 11.49% 4.56% FORMATIONTECH. 85.49% 14.51% MINAGRI 87.95% 4.77% 2.96% 4.32% (Centred'appui formation, EASTA) EAUXET FORETS 88.00% 8.64% 3.36% POPULATION 88.57% 6.20% 5.23% FORCESARMEES 88.92% 9.62% 1.46% ENERGIE,MINES 88.97% 7.47% 3.56% JEUNESSES 90.23% 9.77% ELEVAGE 90.51% 6.81% 2.67% TRANSPORT,METE0 91.34% 6.77% 1.88%(notspecified) SEGENDARMERIE 91.96% 1.14% 6.89% Groupement gendarmerie) BUDGETDEV.PA 93.27% 6.00% 0.72% FONCT.PUBL. 93.36% 1.92% 0.11% 4.61% (not specified) INDARTIS. 93.48% 5.42% 1.10% TRAVAUXPUBL. 94.92% 5.08% COMMERCE 95.26% 2.59% 2.15% MININTER 95.50% 0.39% 4.11% AMENAGEMENTV. 96.22% 1.89% 1.89%(SRD, SRPTF) INFORMATIONS 96.63% 3.37% SE SEC.PUB. 96.69% 3.31% ENVIRONNEMENT 98.09% 1.91% RECHERCHE 98.58% 1.42% PRESIDENCE 99.20% 0.67% 0.13% FINANCEEC. 99.76% 0.07% 0.03% 0.14% (Tresorerie Principale) SEC.PRIVE 99.92% 0.08% PRIMATURE 100% POSTES 100% HCC 100% EINSEGNSUP. 100% A.N.. 100% Total 84.47% 10.65% 0.41% 4.46% Ministries are ranked according to the share o f expenditures assigned below the center, from the highest to the lowest. The level below the region i s ministry-specific and does not necessarily match a geographical level: for example, for the Min Affaires Etrangdres, it identifies the embassies and consulates abroad, for the MinFinance it identifies the TrksoreriePrincipales, etc. Transfers to deconcentrated agencies are here recorded as recurrent expenditures at the central level. 75 Fig 1.TotalExpendituresinEducation,byMinistry 100% 80% 60% 40% 20% 0Yo recurrent investment total Table 2. RecurrentExpendituresbv AdministrativeLevel: Education Non-salary Non-salary exp: disaggregated Total Salary exp exP Purchase of Purchase of Permanent Transfers/ goods services charges subventions Central 129,159,000 101,701,700 27,457,300 7,322,816 12,650,838 1,067,590 6,416,042 31.14% 31.02% 31.57% 23.67% 28.11% 37.35% 78.52% Province 232,000,000 224,649,300 7,350,700 1,559,160 3,959,260 580,180 1,252,180 55.93% 68.53% 8.45% 5.04% 8.80% 20.30% 15.32% Region 118,875 0 118,875 44,288 61,505 13,082 0.03% 0.14% 0.14% 0.14% 0.46% EN1( 1,871,550 120,000 1,751,550 677,650 397,990 173,100 502,810 0.45% 0.04% 2.01% 2.19% 0.88% 6.06% 6.15% District Level: CISCO 30,959,828 1,259,440 29,700,388 7,930,893 21,328,3 13 441,182 7.46% 0.38% 34.15% 25.63% 47.39% 15.43% EPP 16,113,747 0 16,113,747 11,692,807 4,289,614 131,326 3.88% 18.53% 37.79% 9.53% 4.59% Lycee 1,589,760 77,030 1,512,730 646,223 631,250 235,257 0.38% 0.02% 1.74% 2.09% 1.40% 8.23% CEG 2,975,040 2,000 2,973,040 1,067,142 1,689,065 216,833 0.72% 0.00% 3.42% 3.45% 3.75% 7.59% Total 414,787,800 327,809,400 86,978,400 30,940,979 45,007,835 2,858,550 8,171,032 100% 100% 100% 100% 100% 100% 100% 1 .Amounts are expressed in9 =.The % represent the % o fthe total column. The majority (60 percent)of the transfershbventions are transfers to the private sector, the rest is givenbybourses (to ENI)and transfers to organisms public. EN1stands for Ecole National d 'Informatique 76 Table 3. RecurrentEmendituresbv AdministrativeLevel: Health Total Non-salary: disaggregated recurrent T Salary Non-salary Purchase of Purchase of Utilities Transfers/ II goods services (perm.charges) Subvention: Center 01,779,90( 43,437,864 58,342,036 44,111,766 9,791,934 2,407,098 2,03 1,216 45.83% 35.79% 57.94% 74.70% 30.75% 32.56% 84.42% r rovince 88,064,940 76,408,072 11,656,868 4,599,649 5151,174 1331,145 374,900 39.66% 62.96% 11.58% 7.79% 16.17% 20.71% 15.58% egion 761,994 17,843 744,15 1 302,007 344,244 97,900 0.34% 0.01% 0.74% 0.51% 1.08% 1.32% 20,719,860 904,603 19,8 15,257 5,241,261 13,221,612 1,352,384 9.33% 0.75% 19.68% 8.88% 41.51% 18.30% Centres Hosp.District 5,499,760 407,233 5,092,527 2,666,138 1,617,235 809,154 2.48% 0.34% 5.06% 4.51% 5.08% 10.95% 4,057 135,300 3,921,708 1,574,300 1,386,721 960,687 1.83% 0.11% 3.89% 2.67% 4.35% 13% Hopitaux Specialises 1,177,069 48,473 1,128,596 559,869 335,163 233,564 0.53% 0.04% 1.12% 0.95% 1.05% 3.16% L 22,060,500 .21,359,400 100,701,100 59,054,990 31,848,083 7,391,932 2,406,116 100% 100% 100% 100% 100% 100% 100% Source: Tableau de 1 transfersisubventions are transfers to the privatesector. 77 ANNEX 5 THE LOCALFINANCINGGAP METHODOLOGY5' The Approach LocalJinancing gap analysis offers a rapid assessment o f service needs and available financing at the local level. It compares estimates o f local financing needs with available resource flows to facilities for a particular year. This tool i s not designed to measure the degree o f leakage in resource flows, which will be analyzed through separate expenditure tracking surveys (PETS) for the health and education sectors. The approach has three steps: (i) Estimate unmet needs for basic, local services, defined as the difference between a predefinedbenchmark for service delivery and existing local service capacities; (ii) Identzfi financial resources available to support local service delivery, which include financing from local governments, centralized and deconcentrated service providers, donors, NGOs and communities. (iii) Estimate the financing gap by deducting flows o f available financing from estimated needs. Box 1. Calculating the Local Financing Gap Service needs, relative to existing existingjlows of - - I financing gap stocks of services servicefinance InMadagascar, health, education, and water sectors were analyzed because of their significant share o f commune financing, their low degree o f externalities beyond commune boundaries, and their importance as PRSP priorities. Because local financing gap analysis requires intensive empirical field work it is usually only applied to a limited sample. This analysis was based on a national sample o f 15 communes, which were distributed evenly across six provinces and geographical regions, but differ in size and socio-economic characteristics. Questionnaires on budget data, available service infrastructure and administrative capacities were submitted to commune and community representatives as well as to service and project staff at the local, district, and central level. Advantages of local fiscal gap analysis Local fiscal gap analysis requires intensive empirical field work and can therefore only be applied in a limited study sample. Yet, through its strong local research focus it provides important complementary information for other research tools that address service delivery from higher aggregate levels. Specifically it can be applied across sectors and covers all relevant sources o f finance from centralized and deconcentrated expenditures to local counterpart payments by commune governments and communities. 50The local financing gap methodologywill also be presentedin a forthcoming Technical note of the World Bank (see Fengler / Wietzke 2003, PREM-Note). 78 0 adds a local demand-side perspective to more supply-oriented public expenditure reviews andtracking surveys. 0 provides important empirical data on local service finance in countries that lack sufficiently disaggregated expenditure data at the central level. 0 produces relevant information in a decentralization context such as for policy advice on fiscal transfer mechanisms, expenditure and investment planning and local service deliveryinstruments. ESTIMATING LOCAL SERVICE NEEDS Local needs are estimated based on the difference between a predefined level o f local services and existing service levels, as measured by infrastructure, staff and equipment levels. This approach poses methodological challenges because estimated needs depend directly on how service delivery benchmarks are defined. In this study, estimated needs were based on service standards and target populations. Access-based standards were defined, which reflect the quantity and capacity o f local service facilities as well as the range o f services provided. Technical standards were provided by the responsible line ministries or, if unavailable, technical guidelines from donor projects and intemational organizations were used (see Table 1). Needs were defined for those population groups that have access to basic, local services. Local populations without access or specially targeted groups were excluded due to lack o f reliable demographic and household data. Water needs were aggregated at the village level. Table 1. Service Standards" Education Health Water Infrastructure - 1room/ 50 students 12 rooms - 1 htchen / + shower and WC I 1water source per village school healthpost Equipment - 18 tables / class room -12 beads and tables ~ -- 11 text Technical specifications : blackboard -14 chairs and footrests d flow capacity : 0,7M3k book /student and -5 desks and cupboards Diameter drilling : 13 subject -1 stretcher, inches -administrative equipment -5 suspenders Diameter well : >I3 inches Depth o f drilling: 50 2i 60 m medical kits for diagnosis, first Depth o fwell : 15 m aid, small surgery, injection, Maximum time to collect post and prenatal diagnosis, water: 15 minutes child deliverv Staff - 1teacher / 50 students 1medical doctor 1midwife 2 nurses 1sanitary assistant 1guardian Local needs are estimated based on the difference between a hypothetical benchmark o f local service delivery and existing service capacities, as measured by infrastructure, staff and 51Standards in the education sector refer to the minimumneeds estimate. Inthe health sector, all standards refer to Rural Health Centers. Norms for smaller Health posts may vary. 79 equipment levels. Inthe case study a conservative service benchmark was adopted, in order to avoid unrealistically high estimates o f the financing gap. This measure was built around technical standards for physical infrastructure, equipment and staff coverage which were obtained from the responsible line ministries or altematively donor projects (cf. Box for altemative approaches to estimate service needs). In the education sector these standards were applied to the number o f children that were already enrolled in primary schools.52Inthe health and water sector the standards were used to calculate the costs o f a hnctioning health facility at the commune level and a collective potable water system at the village level (excluding individual householdconnections). Box 1.DeterminingLocal Needs Dependingonpolicy objectives, various measures o fneed can be applied Relative needs: This approachtries to identifyregional variations in service supply relative to the national average. This measure i s useful for designing fiscal equalization grants, and other fiscal transfer mechanisms, yet, it tells little about actual needs at the local level. Absolute needs This approach measures absolute needs for service delivery, based on predefinedservice benchmarks or policy objectives (i.e. poverty reduction and social policy perspectives). It i s helpful to redirect policy priorities and expenditure assignments, such as underHIPC andPRSP strategies. The following benchmark indicators can beusedto measure absolute needs: Access based indicators such as Basic Needs Assessments or standard-based indicators that focus on distance to nearest infrastructure facility, capacity or the range o f services provided. Qualitative and outcome-based indictors, such as school repetition and drop-out rates in the education sector, data onnutrition and epidemiological patternsinthe health sector, or user surveys and social assessments that identify specific service needs o f local populations. Coverage: needs estimates can vary considerably, depending on the size o f the target population that i s considered eligible for a specific service. Service coverage could be (i) broadened to include the totality o f the local population, (ii) deepenedto improve services for those parts o f the population that already have access to service facilities, or (iii) target specific beneficiary groups, such as vulnerable households. ANALYZING AVAILABLERESOURCES All commune resources from the center, intermediate service providers, and local sources should be analyzed. InMadagascar, most o f the field work was spent reviewing expenditure data from centralized, deconcentrated or altemative local service providers. Four sources of local financing were reviewed: (i)commune budgets; (ii)central government and deconcentrated service providers; (iii) private contributions by beneficiaries and community groups; (iv) projects, investments and other contributions by donors andNGOs. 52A calculation for total populations inthe schooling age would have been desirable but could not be carried out due to lack o f reliable demographic information. 80 Inthe case study, fixed salary expenditures were excluded from the available resources as these already contributed to existing level of service delivery. In the case of investment and recurrent non-salary expenditures w e assumed that these represented hngible allocations that could b e used to address unmet service needs. CALCULATING THE FINANCING GAP The local financing gap compares unmet local service financing needs to available flows o f service finance. It is calculated by subtracting available resources for service finance from estimated service needs. Inthe case study w e excluded fixed salary expenditures as these already contributed to existing level o f service delivery. In the case o f investment and recurrent non- salary expenditures w e assumed that these represented fungible allocations that could be used to address unmet service needs. 81 ANNEX 6 THE REPRESENTATIVENESS OF THE 232 COMMUNE SAMPLE To account for these constraints a second data set was used, consisting o f a country wide commune census, that was conducted in 2001 by the Agricultural Research Institute (FOFIFA) and Cornel1 University. The commune census has a small section on aggregated expenditures and revenues o f the commune that can be used to verify the representativeness o f the ministry o f Budget sample. In addition, the survey contains a broad range o f indicators on socio-economic characteristics, infrastructure endowments and accessibility o f communes. Table 1. Overlapsbetweenthe Two Samples - Reportedto the ministry ___ Yes Reported in the no I No 105 18 I 123 commune 7.6% 1.3% 8.9% census 76.8% 23.2% I 100% REPRESENTATIVENESSOF THE SAMPLE We assessed the representativeness o f the sample in two steps. First, we use data from the total budget figures o f the commune census to verify whether there are significant differences in the absolute size o f the budget among those communes who report in the census and those who reported their budgets to the central ministry. Second, we looked at whether the likelihood o f reporting the budget is systematically related to some structural characteristics o f the commune, such as the population and the degree o f isolation o f a commune. Using data from the Commune Census, on average, we find no significant differences in the budget size o f those communes who report the budget to the ministry o f budget and those who do not do so. Communes who are reporting their budget are also on average smaller, even if the difference i s not significant. Table - below suggests that our sample i s not skewed towards communes with larger budgets. However, while the variability in reporting i s not related to the size o f the budget, it seems to be related to structural constraints. Communes that are more remote and less accessible53are less likely to report the budget. Moreover, neither the smaller nor the biggest communes, in terms o f population size are likely to send their budget o f the 53Inparticular, we find that communes that are more distance to urban centers or that do not have a bush taxi connection-are less likely to report the budget. 82 ministry.54This leads us to conclude that the sample of 321 communes is more representative for less isolated as well as for medium rural communes.55 Table 2. OverlapbetweenReportedBudgetswith Budgetsi the CommuneCensus COMMUNE CENSUS M I N I S T R Y o f BUDGET Whole Non Reporting difference Reporting difference census reporting budget replrep budget MinBudg. Census- budget (p value) -__________ (p value) Tot. revenue 96,192.8 98,076.7 90,205.5 0.60 n.s. 86,410.9 0.52 n.s. (232,400) (262,300) (84,261) (60,502) Share 0.71 0.72 0.68 0.31 n.s. 0.76 o.oo** transfedtot (0.48) (0.53) (0.24) (0.16) revenue Tot. 81,310.8 83,260.2 75,122.6 0.41 74,272.6 0.85 n.s. expenditures (150,200) (168,100) (66,230) n.s. (47,099) Investment 19,876.2 20,245.1 18,689.5 0.55 n.s. 12,183.4 0.02** expenditures (37,677) (36,765) (40,524) (15,897) Recurrent 63,838.7 65,643.8 58,126.5 0.38 n.s. 66,620.5 0.07* expenditures (132,600) (150,600) (38,262) - (407436)-----. __I_- Population 15026.4 15454.7 13609.3 0.36 (850.9) (1096.7) (519.9) Time to 19.00 20.05 15.52 O.OO** nearest urban (0.62) (0.73) (1.15) center (hrs) Note: figures are in 1000 FMG. S G rd errors are in parentheses. (n.s. means statistically not significant at 5 percent confidence level). Lowess smwther. bandwidth = .0 -1.906384 1 0 Tailie de50000 rooobo la population odds reportingbudgetvs population size 54The fact that the likelihood o f reporting the budget decreases with population size indicates that there are few urban communes in our sample. Our discussion o f the case o f urban communes will therefore be limited to the analysis o f aggregate budget data. 55Future work will check how this selectivity affects the main correlations we find between revenue/expenditure composition and the socio-economic characteristics o f the communes. 83 -.792533r,--.I Lowess smoother, bandwidth = .6 -6.06972 -__ 0 50 100 7-1 150 200 odds reporting budget vs time to nearest urban center time to the nearest town (hrs) 84 ANNEX 7 LOCAL GOVERNMENTORGANIGRAM PROVINCE : TOAMASINA DISTRICT : Brickavile COMMUNE :Andovoranto (2nd Category) CONSEIL COMMUNAL 1Vice-PrCsident 2 Rapporteurs 10Membres 1Maire FINANCES ETAT-CIVIL 1 ADMINISTRATION GENERALE 1SecrCtaire - TrCsorier 1SecrCtaire EC 1Gardien 14Conseillers communaux .... 5 membres du personnel, dont : 1Maire Superficie : 325km2 1Adjoint Population : 11291 Hab. 1SecrCtaire - TrCsorier 1SecrCtaire Etat-Civil Densitt : 35 H a b . h z 1Gardien 1Dispensateur et 1Gardien CSB. 85 ANNEX 8 LOCAL GOVERNMENTREVENUEASSIGNMENTS Table ,Tax Assignments Tax level Assignment Recovery mechanism Determinedby Commune 100%Commune Treasury council Treasury Treasury council 15%o fprincipal 15% Region Treasury 15%Commune Revenue registration duties Administration Additionnelle 1'Enregistrementdes mutationsdes biens) 8. Taxes on Licenses for Art. 03.02.11 Variable accordingto 100%Commune DAA sale of alcoholic beverages 03.02.14 licences andpopulation 03.02.22 03.02.26 du CGI (5.000/2.000 FMGper 1999 License of 12hours) Art. 10.06.12 to 10.06.91 9. Tax on slot machines Art. 03.03.12 duCGI 100%Commune DAA 1999 10. Tax on gambling Art. 03.02.12 Revenue Art. 10.07.01a Adrmnistration 10.07.07 11.Tax on transactions Art. 06.02.01ff Revenue (Taxe Unique sur les Administration transactions, TST) I 13. User fee for waste Art. 01.09.05 du CGI Determinedby commune 100%Commune management 1999 Zouncil Redevance de collecte et traitement des Ordures 86 I Table 2. Administrativefees Fee I Collection 1 Execution o f charts (chartres) jCommune Fee for Secrttaires d'ttat civil /Commune Table3. User fees I Collection fee (mining.,naturalresources) IFokontanv I IIRent from communal housing/ infrastructure ICommune TransDort rights (droit de voirie) lcommune I1 87 ANNEX 9 BIBLIOGRAPHY African Development Bank. 2001, Gestion des Finances Publiques dans un Cadre Dkentralise. Contribution a l'klaboration et lamise en ceuvre de la stratkgie de Reduction de lapauvrete, Abidjan. Akin, John& Hutchinson, Paul& Strumpf, Koleman. 2001. "Decentralization andGovernment Provision o f Public Goods: The Public Health Sector inUganda" (Draft), Measure Evaluation, Chapel Hill [ http://www.cpc.unc.edu/measure/publications/workingpapers/wpOl35.pdfl. Bahl, Roy. 2002. "The Quest for LocalRevenues: Income & Property Taxes; presentation at an ASEMworkshop inBali, January 10-11 [http://www.worldbank.org/ wbi/publicfinance/documents/ASEM/presentations/bahl_income.pdfl. Bahl,Roy. 1999. "Implementation Rulesfor FiscalDecentralization". Intemational Studies Program, School o fPolicy Studies, Georgia State University,Atlanta. Bahl, Roy. 1994. "Revenues and Revenue Assignment:IntergovernmentalFiscal relations inthe Russian Federation," in:Russia and the Challenge o fFiscal Federalismpp.129-180, ed. Christine Wallich, Washington, D.C.: World Bank. Bardhan, Pranab. 2002a. "Decentralization and Govemance". Joumal o f Economic Perspectives XX. Bardhan, Pranab. 2002b. "Decentralization o f Govemance and Development". In:Joumal o f Economic Perspectives, Fall 2002. Bardhan, P. and D.Mookherjee. 2000. "Relative Capture o f Govemance at Local andNational Levels", American Economic Review, vol. 90(2). Barrientos, Jorge C. 1999. Coordinating Poverty Alleviation ProFiramswith Regional and Local Governments: The Experience o fthe Chilean Social Fund(FOSIS), Social ProtectionDiscussion PaperNo. 9933. Bird, Richard. 2000. "Intergovernmental Fiscal Relations inLatinAmerica: PolicyDesign and Policy Outcomes". Sustainable Development Department Technical Papers Series, Inter- American Development Bank, Sustainable Development Department, Social Development Division, Washington D.C. Brown, Mervyn. 1995. "A History o f Madagascar". MarkusWiener Publishers, Princeton. Brunet,Franqoise & DechenSeheno, Boris. 2003. "Etude sur les Systkmes de Transfert en Faveur des Pauvres". Secretariat Technique de 1'Ajustement, PAIGEP. 88 Brosio, Giorgio. 2000. "Decentralization inAfrica". InternationalMonetary Fund. Burki, Javed & Perry, Guillermo E. 1999. "Accountability o fthe Public Sector." WorldBank Latin American and Caribbean Studies, Washington D.C.: World Bank. Burki,Javed; Perry, Guillermo & Dillinger, William. 1999. "Beyond the Center -Decentralizing the State." World Bank Latin American and Caribbean Studies, Washington D.C.: World Bank. Collins, Paul. 2003. "Editorial". Public Administration and Development, Wiley Interscience, London. Crook, Richard C. 2003. "Decentralization and Poverty Reduction inAfrica: the Politics o f Local-CentralRelations". Public Administration and Development, Wiley Interscience, Brighton. FEW/ UNDCF. 2003. Projet de Developpement Regional d'Ambato-Boeny (PDRAP).Aquis et Leqons de 1))experience d'appui Au Developpement Local. Frank, Jonas. 2002. Decentralization to the Regional Level in Latin America, Cases of Venezuela, Colombia, Ecuador, Peru, Bolivia and Chile, With References to Spain and France, Draft for Presentation, World Bank, Washington DC. Fuhr,Harald. 1999. "Institutional Change andNew Incentive Structures for Development: Can Decentralization and Better Local Governance Help?" Gazetim-Panjakan'ny Repoblikan'I Madagasikara. (2002). Journal Officiel de la Republique de Madagascar. Madagascar. Government o fMadagascar (2001): Etude en Vue de 1'AmClioration du Fonctionnement et du Renforcement des CapacitCsdes Groupes de Travail de DCveloppement Regional, GTDR, Antananarivo. Helmsing, A.H.J. 2003. "Local Economic Development: New Generations of Actors, Policies and Instrumentsfor Africa". Public Administration and Development, Wiley Interscience, The Hague. Kahkonen, Satu. 2001. "Decentralization and Governance: does Decentralization improve Public Service Delivery?"PREMnote 55, Poverty ReductionandEconomic Management Network, Washington D.C.: World Bank, King, ElisabethM.& Ozler, Berk. YEAR. "What's Decentralization Got To Do With Learning? The Case o f Nicaragua's School Autonomy Reform", Development Research Group, Washington D.C.: World Bank. Knack, Stephen& Manning, Nick. 2000. "Toward more Operationally Relevant Indicators o f Governance". PREMnote 49, Poverty Reduction and Economic Management Network, Washington D.C.: World Bank. 89 Lienert, Ian& Paoletti, Marc. 2001. "Madagascar: Execution Budgktaire - L e Tableau de Bord." Fiscal Affairs Department, IntemationalMonetary Fund, Washington D.C. Litvack, Jennie & Seddon, Jessica. 1999. DecentralizationBriefing Notes. WBI Working Papers, Washington D.C.: The World Bank Institute. Litvack, Jennie; Ahmad, Junaid & Bird, Richard. 1998. "Rethinking Decentralization in Developing Countries". Sector Studies Series, Poverty Reduction and Economic Management, Washington D.C.: World Bank. Manor, James. 1999. "The Political Economy o f Democratic Decentralization". Directions in Development, Washington D.C.: World Bank. Mira Mita. 2003. "Etude sur les besoins financiers des Communes a Madagascar". Mistiaen, J.; Ozler, B.; Razafimanantena, T. & Razafindravonona,J. 2002. "PuttingWelfare on the Map inMadagascar". Africa RegionWorking Paper Series, No.34, Washington D.C.: World Bank. Ndegwa, Stephen N.2002. "Decentralization inAfrica: A Stocktaking Survey". Africa Region Working Paper Series No. 40, Washington D.C.: World Bank. Oates, Wallace E. 1999. "An Essay on FiscalFederalism" in: Journal o fEconomic Literature N o 37. Olowu, Delewe. 2001. Decentralization Policies and Practices under Structural Adjustment and Democratization in Africa, UNRISD, Democracy, Governance and Human Rights Program Paper N o 4. Parker, Andrew & Rodrigo Serrano. 2000. Promoting Good Local Governance through Social Funds and Decentralization, Social ProtectionDiscussionPaper No. 0022; Projet de 1'Ambassade de Suisse a Madagascar et de 1'Action de CarCme Suisse (AdC) (2003) : Appui A la Gouvemance, la dkcentralisation et la democratie, Rapport de la premibre phase: Janvier A avril2003 Prud'homme, Rkmy. 2003. "Fiscal DecentralizationinAfrica: a Framework for Considering Reform". Public Administration and Development, Wiley Interscience, Paris. Ravallion, Martin. 1998. "Reaching Poor Areas ina Federal System". World Bank Research Working Paper, Washington D.C.: World Bank. Reinnika, Ritva & Svensson, Jacob. 2001. "Explaining Leakage o fPublic Funds" World Bank Policy Research Working Paper, Washington D.C.: World Bank. Rodden, Jonathan. 2000. "Decentralization and the Challenge o f HardBudget Constraints". PREMnote 41, Poverty Reduction and Economic Management Network, Washington D.C.: World Bank. Rondinelli, Dennis A. 1999. "What i s Decentralization?" Decentralization BriefingNotes edited by: Jennie Litvack andJessica Seddon, Washington D.C.: World Bank Institute. 90 Seheno, Boris D. 1997. "Travaux Juridiques sur la Dkcentralisation. Madagascar". Shah, Anwar. 2001. "Fiscal DecentralizationinDeveloping and Transition Economies: Progress, Problems and the Promise". Paper presentedat the Universityo f St. Gallen. Shah, Anwar.1994. "The Reform o f IntergovemmentalFiscal RelationsinDeveloping and EmergingMarket Economies". Policy and Research Series No. 23, Washington D.C.: World Bank. Smoke, Paul. 2001. "Fiscal DecentralizationinDeveloping Countries: A Review o f Current Concepts and Practice". United Nations Research Institute for Social Development, Geneva. Smoke, Paul. 2003. "Decentralization inAfrica: Goals, Dimensions, Myths and Challenges.'' Public Administration and Development, Wiley Interscience, New York. UnitedNations Development Programme. 2000. Rapport national sur le dkveloppement humain Madagascar. Antananarivo. UnitedNations Development Programme. 2002. "Local Governance for Poverty Reduction in Africa". FifthAfrica Governance Forum, UnitedNations Economic Commission for Africa World Bank. 1996. "Vietnam Fiscal Decentralization and the Delivery o fRural Services". Economic Report No. 15745-VN, Country Operations Division, Country Department I, East Asia and Pacific Region, Washington D.C. World Bank. 1997. "The State ina Changing World". World Development Report 1997, New York: Oxford University Press. World Bank. 1998. "Decentralization inMadagascar: a Strategy for Successful Implementation", Washington D.C. World Bank. 1999. "Entering the 21StCentury". World Development Report 2000, New York: OxfordUniversityPress. World Bank. 2001, "The UnitedRepublic o fTanzania. Decentralization Sector Work", Washington D.C. World Bank.2002. "Education and Training inMadagascar. Towards a Policy Agenda for Economic Growth and Poverty Reduction". Report No. 22389-MAG, Washington D.C. 91 ANNEX 10. GLOSSARY OF FRENCH AND MALAGASYTERMINOLOGY Canton :Administrative unit regrouping several communes duringthe colonial period Circonscription FinanciBre (CF) :Administrative unit o f Financial organization CentreFiscal :Fiscal Center Centre Sanitaire de Base (CSB) :Communal level Health Center Comite'Local de De'veloppement (CLD) :Local Development Commitee Commission Centrale des Marche's (CCM) :Public Procurement Committee Comptes Administratifs :BudgetAccounts ContrGle des De'pensesEngage'es (CDE) :Internal State Control Organ, incharge o f ex-ante controls o f expenditure procedures inline ministries andpublic administrations De'le'gue'Administratifd 'Arrondissement (DAA) :LocalAdministrative Delegate - at the Commune and/or Neighborhood level (inurbancommunes) De'le'gue' Ge'ne'raldu Gouvernement (DGG) :State Representative at the Province level (not functional at this stage) De'partement :State level administrative division headedby a souspre'fet. De'pute' :Member o fParliament (National Assembly) Etat Civil :Civil Registry Fartany : Former state level administrative unit (equivalent o fprovinces) Firaisana :Former local level administrative unit (equivalent o f communes ) Fivondrona :District Fokonolona :traditional unit o f the Malagasy community Fokontany :semi-formal administrativeunit at village and/ or neighborhood level Gestionnaire :Civil servant or elected representative incharge o f managing public funds Inspecteur Ge'ne'ralde 1'Etat (IGE) :Central auditing body - Administrative Inspection 92 Ordonnateur :Higher level civil servant who orders commitment and payment o fpublic finds Plan Communal deDe'veloppement(PCD) :Municipal DevelopmentPlan Pre'fecture :Deconcentrated administrative unit (about equivalent o fthe regional level) Pre'fet : Deconcentrated state representative representing all ministers individually and collectively at the "regional" level. Thepre'fet is appointed by a Presidential decree and reports directly to the Ministry o f Interior. H e i s also incharge o f safety and security issues. Pre'sident de la De'le'gationSpe'ciale(PDS) :State representative nominated centrally to manage provincial affairs Secre'taired 'Etat Civil :Communal agent incharge o fmaintainingthe Civil registry Sous-pre'fecture :Deconcentrated administrativeunit, equivalent to the department level Sous-pre'fet :Deconcentrated state representative, head o fthe sous-pre'fecture Tutelle :Authority for control and oversight (delegated to the Prbfet ineach Department). ZoneAdministrative Pe'dagogique:Communal level Administrative unit for Education 93